1.Ditto: Building Digital Twins of Articulated Objects from Interaction ⬇️
Digitizing physical objects into the virtual world has the potential to unlock new research and applications in embodied AI and mixed reality. This work focuses on recreating interactive digital twins of real-world articulated objects, which can be directly imported into virtual environments. We introduce Ditto to learn articulation model estimation and 3D geometry reconstruction of an articulated object through interactive perception. Given a pair of visual observations of an articulated object before and after interaction, Ditto reconstructs part-level geometry and estimates the articulation model of the object. We employ implicit neural representations for joint geometry and articulation modeling. Our experiments show that Ditto effectively builds digital twins of articulated objects in a category-agnostic way. We also apply Ditto to real-world objects and deploy the recreated digital twins in physical simulation. Code and additional results are available at this https URL
2.Less is More: Surgical Phase Recognition from Timestamp Supervision ⬇️
Surgical phase recognition is a fundamental task in computer-assisted surgery systems. Most existing works require expensive frame-wise annotations, which is very time-consuming. In this paper, we introduce timestamp supervision to surgical phase recognition for the first time, which only requires randomly labeling one frame for each phase in a video. With timestamp supervision, current methods in natural videos aim to generate pseudo labels of full frames. However, due to the surgical videos containing ambiguous boundaries, these methods would generate many noisy and inconsistent pseudo labels, leading to limited performance. We argue that less is more in surgical phase recognition,~\ie, less but discriminative pseudo labels outperform full but ambiguous frames. To this end, we propose a novel method called uncertainty-aware temporal diffusion to generate trustworthy pseudo labels. Our approach evaluates the confidence of generated pseudo labels based on uncertainty estimation. Then, we treat the annotated frames as anchors and make pseudo labels diffuse to both sides, starting from anchors and stopping at the high-uncertainty frames. In this way, our proposed method can generate contiguous confident pseudo labels while discarding the uncertain ones. Extensive experiments demonstrate that our method not only significantly save annotation cost, but also outperforms fully supervised methods. Moreover, our proposed approach can be used to clean noisy labels near boundaries and improve the performance of the current surgical phase recognition methods.
3.Flexible-Modal Face Anti-Spoofing: A Benchmark ⬇️
Face anti-spoofing (FAS) plays a vital role in securing face recognition systems from presentation attacks. Benefitted from the maturing camera sensors, single-modal (RGB) and multi-modal (e.g., RGB+Depth) FAS has been applied in various scenarios with different configurations of sensors/modalities. Existing single- and multi-modal FAS methods usually separately train and deploy models for each possible modality scenario, which might be redundant and inefficient. Can we train a unified model, and flexibly deploy it under various modality scenarios? In this paper, we establish the first flexible-modal FAS benchmark with the principle `train one for all'. To be specific, with trained multi-modal (RGB+Depth+IR) FAS models, both intra- and cross-dataset testings are conducted on four flexible-modal sub-protocols (RGB, RGB+Depth, RGB+IR, and RGB+Depth+IR). We also investigate prevalent deep models and feature fusion strategies for flexible-modal FAS. We hope this new benchmark will facilitate the future research of the multi-modal FAS. The protocols and codes are available at this https URL.
4.Self-Supervised Class-Cognizant Few-Shot Classification ⬇️
Unsupervised learning is argued to be the dark matter of human intelligence. To build in this direction, this paper focuses on unsupervised learning from an abundance of unlabeled data followed by few-shot fine-tuning on a downstream classification task. To this aim, we extend a recent study on adopting contrastive learning for self-supervised pre-training by incorporating class-level cognizance through iterative clustering and re-ranking and by expanding the contrastive optimization loss to account for it. To our knowledge, our experimentation both in standard and cross-domain scenarios demonstrate that we set a new state-of-the-art (SoTA) in (5-way, 1 and 5-shot) settings of standard mini-ImageNet benchmark as well as the (5-way, 5 and 20-shot) settings of cross-domain CDFSL benchmark. Our code and experimentation can be found in our GitHub repository: this https URL.
5.Bias in Automated Image Colorization: Metrics and Error Types ⬇️
We measure the color shifts present in colorized images from the ADE20K dataset, when colorized by the automatic GAN-based DeOldify model. We introduce fine-grained local and regional bias measurements between the original and the colorized images, and observe many colorization effects. We confirm a general desaturation effect, and also provide novel observations: a shift towards the training average, a pervasive blue shift, different color shifts among image categories, and a manual categorization of colorization errors in three classes.
6.FUN-SIS: a Fully UNsupervised approach for Surgical Instrument Segmentation ⬇️
Automatic surgical instrument segmentation of endoscopic images is a crucial building block of many computer-assistance applications for minimally invasive surgery. So far, state-of-the-art approaches completely rely on the availability of a ground-truth supervision signal, obtained via manual annotation, thus expensive to collect at large scale. In this paper, we present FUN-SIS, a Fully-UNsupervised approach for binary Surgical Instrument Segmentation. FUN-SIS trains a per-frame segmentation model on completely unlabelled endoscopic videos, by solely relying on implicit motion information and instrument shape-priors. We define shape-priors as realistic segmentation masks of the instruments, not necessarily coming from the same dataset/domain as the videos. The shape-priors can be collected in various and convenient ways, such as recycling existing annotations from other datasets. We leverage them as part of a novel generative-adversarial approach, allowing to perform unsupervised instrument segmentation of optical-flow images during training. We then use the obtained instrument masks as pseudo-labels in order to train a per-frame segmentation model; to this aim, we develop a learning-from-noisy-labels architecture, designed to extract a clean supervision signal from these pseudo-labels, leveraging their peculiar noise properties. We validate the proposed contributions on three surgical datasets, including the MICCAI 2017 EndoVis Robotic Instrument Segmentation Challenge dataset. The obtained fully-unsupervised results for surgical instrument segmentation are almost on par with the ones of fully-supervised state-of-the-art approaches. This suggests the tremendous potential of the proposed method to leverage the great amount of unlabelled data produced in the context of minimally invasive surgery.
7.When Did It Happen? Duration-informed Temporal Localization of Narrated Actions in Vlogs ⬇️
We consider the task of temporal human action localization in lifestyle vlogs. We introduce a novel dataset consisting of manual annotations of temporal localization for 13,000 narrated actions in 1,200 video clips. We present an extensive analysis of this data, which allows us to better understand how the language and visual modalities interact throughout the videos. We propose a simple yet effective method to localize the narrated actions based on their expected duration. Through several experiments and analyses, we show that our method brings complementary information with respect to previous methods, and leads to improvements over previous work for the task of temporal action localization.
8.Using Navigational Information to Learn Visual Representations ⬇️
Children learn to build a visual representation of the world from unsupervised exploration and we hypothesize that a key part of this learning ability is the use of self-generated navigational information as a similarity label to drive a learning objective for self-supervised learning. The goal of this work is to exploit navigational information in a visual environment to provide performance in training that exceeds the state-of-the-art self-supervised training. Here, we show that using spatial and temporal information in the pretraining stage of contrastive learning can improve the performance of downstream classification relative to conventional contrastive learning approaches that use instance discrimination to discriminate between two alterations of the same image or two different images. We designed a pipeline to generate egocentric-vision images from a photorealistic ray-tracing environment (ThreeDWorld) and record relevant navigational information for each image. Modifying the Momentum Contrast (MoCo) model, we introduced spatial and temporal information to evaluate the similarity of two views in the pretraining stage instead of instance discrimination. This work reveals the effectiveness and efficiency of contextual information for improving representation learning. The work informs our understanding of the means by which children might learn to see the world without external supervision.
9.Typography-MNIST (TMNIST): an MNIST-Style Image Dataset to Categorize Glyphs and Font-Styles ⬇️
We present Typography-MNIST (TMNIST), a dataset comprising of 565,292 MNIST-style grayscale images representing 1,812 unique glyphs in varied styles of 1,355 Google-fonts. The glyph-list contains common characters from over 150 of the modern and historical language scripts with symbol sets, and each font-style represents varying subsets of the total unique glyphs. The dataset has been developed as part of the CognitiveType project which aims to develop eye-tracking tools for real-time mapping of type to cognition and to create computational tools that allow for the easy design of typefaces with cognitive properties such as readability. The dataset and scripts to generate MNIST-style images for glyphs in different font styles are freely available at this https URL.
10.Paraphrasing Magritte's Observation ⬇️
Contrast Sensitivity of the human visual system can be explained from certain low-level vision tasks (like retinal noise and optical blur removal), but not from others (like chromatic adaptation or pure reconstruction after simple bottlenecks). This conclusion still holds even under substantial change in stimulus statistics, as for instance considering cartoon-like images as opposed to natural images (Li et al. Journal of Vision, 2022, Preprint arXiv:2103.00481).
In this note we present a method to generate original cartoon-like images compatible with the statistical training used in (Li et al., 2022). Following the classical observation in (Magritte, 1929), the stimuli generated by the proposed method certainly are not what they represent: Ceci n'est pas une pipe. The clear distinction between representation (the stimuli generated by the proposed method) and reality (the actual object) avoids eventual problems for the use of the generated stimuli in academic, non-profit, publications.
11.Reversible data hiding with dual pixel-value-ordering and1minimum prediction error expansion ⬇️
Pixel Value Ordering (PVO) holds an impressive property for high fidelity Reversible Data Hiding (RDH). In this paper, we introduce a dual-PVO (dPVO) for Prediction Error Expansion(PEE), and thereby develop a new RDH scheme to offer a better rate-distortion performance. Particularly, we propose to embed in two phases: forward and backward. In the forward phase, PVO with classic PEE is applied to every non-overlapping image block of size 1x3. In the backward phase,minimum-set and maximum-set of pixels are determined from the pixels predicted in the forward phase. The minimum set only contains the lowest predicted pixels and the maximum set contains the largest predicted pixels of each image block. Proposed dPVO withPEE is then applied to both sets, so that the pixel values of the minimum set are increased and that of the maximum set are decreased by a unit value. Thereby, the pixels predicted in the forward embedding can partially be restored to their original values resulting in both better-embedded image quality and a higher embedding rate. Experimental results have recorded a promising rate-distortion performance of our scheme with a significant improvement of embedded image quality at higher embedding rates compared to the popular and state-of-the-art PVO-based RDHschemes.
12.Learning to Adapt to Light ⬇️
Light adaptation or brightness correction is a key step in improving the contrast and visual appeal of an image. There are multiple light-related tasks (for example, low-light enhancement and exposure correction) and previous studies have mainly investigated these tasks individually. However, it is interesting to consider whether these light-related tasks can be executed by a unified model, especially considering that our visual system adapts to external light in such way. In this study, we propose a biologically inspired method to handle light-related image-enhancement tasks with a unified network (called LA-Net). First, a frequency-based decomposition module is designed to decouple the common and characteristic sub-problems of light-related tasks into two pathways. Then, a new module is built inspired by biological visual adaptation to achieve unified light adaptation in the low-frequency pathway. In addition, noise suppression or detail enhancement is achieved effectively in the high-frequency pathway regardless of the light levels. Extensive experiments on three tasks -- low-light enhancement, exposure correction, and tone mapping -- demonstrate that the proposed method almost obtains state-of-the-art performance compared with recent methods designed for these individual tasks.
13.Learning to Detect People on the Fly: A Bio-inspired Event-based Visual System for Drones ⬇️
We demonstrate for the first time that a biologicallyplausible spiking neural network (SNN) equipped with Spike- Timing-Dependent Plasticity (STDP) learning can continuously learn to detect walking people on the fly using retina-inspired, event-based camera data. Our pipeline works as follows. First, a short sequence of event data (< 2 minutes), capturing a walking human from a flying drone, is shown to a convolutional SNNSTDP system which also receives teacher spiking signals from a convolutional readout (forming a semi-supervised system). Then, STDP adaptation is stopped and the learned system is assessed on testing sequences. We conduct several experiments to study the effect of key mechanisms in our system and we compare our precision-recall performance to conventionally-trained CNNs working with either RGB or event-based camera frames.
14.360 Depth Estimation in the Wild -- The Depth360 Dataset and the SegFuse Network ⬇️
Single-view depth estimation from omnidirectional images has gained popularity with its wide range of applications such as autonomous driving and scene reconstruction. Although data-driven learning-based methods demonstrate significant potential in this field, scarce training data and ineffective 360 estimation algorithms are still two key limitations hindering accurate estimation across diverse domains. In this work, we first establish a large-scale dataset with varied settings called Depth360 to tackle the training data problem. This is achieved by exploring the use of a plenteous source of data, 360 videos from the internet, using a test-time training method that leverages unique information in each omnidirectional sequence. With novel geometric and temporal constraints, our method generates consistent and convincing depth samples to facilitate single-view estimation. We then propose an end-to-end two-branch multi-task learning network, SegFuse, that mimics the human eye to effectively learn from the dataset and estimate high-quality depth maps from diverse monocular RGB images. With a peripheral branch that uses equirectangular projection for depth estimation and a foveal branch that uses cubemap projection for semantic segmentation, our method predicts consistent global depth while maintaining sharp details at local regions. Experimental results show favorable performance against the state-of-the-art methods.
15.Planckian jitter: enhancing the color quality of self-supervised visual representations ⬇️
Several recent works on self-supervised learning are trained by mapping different augmentations of the same image to the same feature representation. The set of used data augmentations is of crucial importance for the quality of the learned feature representation. We analyze how the traditionally used color jitter negatively impacts the quality of the color features in the learned feature representation. To address this problem, we replace this module with physics-based color augmentation, called Planckian jitter, which creates realistic variations in chromaticity, producing a model robust to llumination changes that can be commonly observed in real life, while maintaining the ability to discriminate the image content based on color information. We further improve the performance by introducing a latent space combination of color-sensitive and non-color-sensitive features. These are found to be complementary and the combination leads to large absolute performance gains over the default data augmentation on color classification tasks, including on Flowers-102 (+15%), Cub200 (+11%), VegFru (+15%), and T1K+ (+12%). Finally, we present a color sensitivity analysis to document the impact of different training methods on the model neurons and we show that the performance of the learned features is robust with respect to illuminant variations.
16.Unified smoke and fire detection in an evolutionary framework with self-supervised progressive data augment ⬇️
Few researches have studied simultaneous detection of smoke and flame accompanying fires due to their different physical natures that lead to uncertain fluid patterns. In this study, we collect a large image data set to re-label them as a multi-label image classification problem so as to identify smoke and flame simultaneously. In order to solve the generalization ability of the detection model on account of the movable fluid objects with uncertain shapes like fire and smoke, and their not compactible natures as well as the complex backgrounds with high variations, we propose a data augment method by random image stitch to deploy resizing, deforming, position variation, and background altering so as to enlarge the view of the learner. Moreover, we propose a self-learning data augment method by using the class activation map to extract the highly trustable region as new data source of positive examples to further enhance the data augment. By the mutual reinforcement between the data augment and the detection model that are performed iteratively, both modules make progress in an evolutionary manner. Experiments show that the proposed method can effectively improve the generalization performance of the model for concurrent smoke and fire detection.
17.ActionFormer: Localizing Moments of Actions with Transformers ⬇️
Self-attention based Transformer models have demonstrated impressive results for image classification and object detection, and more recently for video understanding. Inspired by this success, we investigate the application of Transformer networks for temporal action localization in videos. To this end, we present ActionFormer -- a simple yet powerful model to identify actions in time and recognize their categories in a single shot, without using action proposals or relying on pre-defined anchor windows. ActionFormer combines a multiscale feature representation with local self-attention, and uses a light-weighted decoder to classify every moment in time and estimate the corresponding action boundaries. We show that this orchestrated design results in major improvements upon prior works. Without bells and whistles, ActionFormer achieves 65.6% mAP at tIoU=0.5 on THUMOS14, outperforming the best prior model by 8.7 absolute percentage points and crossing the 60% mAP for the first time. Further, ActionFormer demonstrates strong results on ActivityNet 1.3 (36.0% average mAP) and the more recent EPIC-Kitchens 100 (+13.5% average mAP over prior works). Our code is available at this http URL
18.Practical Network Acceleration with Tiny Sets ⬇️
Network compression is effective in accelerating the inference of deep neural networks, but often requires finetuning with all the training data to recover from the accuracy loss. It is impractical in some applications, however, due to data privacy issues or constraints in compression time budget. To deal with the above issues, we propose a method named PRACTISE to accelerate the network with tiny sets of training images. By considering both the pruned part and the unpruned part of a compressed model, PRACTISE alleviates layer-wise error accumulation, which is the main drawback of previous methods. Furthermore, existing methods are confined to few compression schemes, have limited speedup in terms of latency, and are unstable. In contrast, PRACTISE is stable, fast to train, versatile to handle various compression schemes, and achieves low latency. We also propose that dropping entire blocks is a better way than existing compression schemes when only tiny sets of training data are available. Extensive experiments demonstrate that PRACTISE achieves much higher accuracy and more stable models than state-of-the-art methods.
19.PCRP: Unsupervised Point Cloud Object Retrieval and Pose Estimation ⬇️
An unsupervised point cloud object retrieval and pose estimation method, called PCRP, is proposed in this work. It is assumed that there exists a gallery point cloud set that contains point cloud objects with given pose orientation information. PCRP attempts to register the unknown point cloud object with those in the gallery set so as to achieve content-based object retrieval and pose estimation jointly, where the point cloud registration task is built upon an enhanced version of the unsupervised R-PointHop method. Experiments on the ModelNet40 dataset demonstrate the superior performance of PCRP in comparison with traditional and learning based methods.
20.Reducing Overconfidence Predictions for Autonomous Driving Perception ⬇️
In state-of-the-art deep learning for object recognition, SoftMax and Sigmoid functions are most commonly employed as the predictor outputs. Such layers often produce overconfident predictions rather than proper probabilistic scores, which can thus harm the decision-making of `critical' perception systems applied in autonomous driving and robotics. Given this, the experiments in this work propose a probabilistic approach based on distributions calculated out of the Logit layer scores of pre-trained networks. We demonstrate that Maximum Likelihood (ML) and Maximum a-Posteriori (MAP) functions are more suitable for probabilistic interpretations than SoftMax and Sigmoid-based predictions for object recognition. We explore distinct sensor modalities via RGB images and LiDARs (RV: range-view) data from the KITTI and Lyft Level-5 datasets, where our approach shows promising performance compared to the usual SoftMax and Sigmoid layers, with the benefit of enabling interpretable probabilistic predictions. Another advantage of the approach introduced in this paper is that the ML and MAP functions can be implemented in existing trained networks, that is, the approach benefits from the output of the Logit layer of pre-trained networks. Thus, there is no need to carry out a new training phase since the ML and MAP functions are used in the test/prediction phase.
21.RNGDet: Road Network Graph Detection by Transformer in Aerial Images ⬇️
Road network graphs provide critical information for autonomous vehicle applications, such as motion planning on drivable areas. However, manually annotating road network graphs is inefficient and labor-intensive. Automatically detecting road network graphs could alleviate this issue, but existing works are either segmentation-based approaches that could not ensure satisfactory topology correctness, or graph-based approaches that could not present precise enough detection results. To provide a solution to these problems, we propose a novel approach based on transformer and imitation learning named RNGDet (\underline{R}oad \underline{N}etwork \underline{G}raph \underline{Det}ection by Transformer) in this paper. In view of that high-resolution aerial images could be easily accessed all over the world nowadays, we make use of aerial images in our approach. Taken as input an aerial image, our approach iteratively generates road network graphs vertex-by-vertex. Our approach can handle complicated intersection points of various numbers of road segments. We evaluate our approach on a publicly available dataset. The superiority of our approach is demonstrated through the comparative experiments.
22.Cross-view and Cross-domain Underwater Localization based on Optical Aerial and Acoustic Underwater Images ⬇️
Cross-view image matches have been widely explored on terrestrial image localization using aerial images from drones or satellites. This study expands the cross-view image match idea and proposes a cross-domain and cross-view localization framework. The method identifies the correlation between color aerial images and underwater acoustic images to improve the localization of underwater vehicles that travel in partially structured environments such as harbors and marinas. The approach is validated on a real dataset acquired by an underwater vehicle in a marina. The results show an improvement in the localization when compared to the dead reckoning of the vehicle.
23.Applying adversarial networks to increase the data efficiency and reliability of Self-Driving Cars ⬇️
Convolutional Neural Networks (CNNs) are vulnerable to misclassifying images when small perturbations are present. With the increasing prevalence of CNNs in self-driving cars, it is vital to ensure these algorithms are robust to prevent collisions from occurring due to failure in recognizing a situation. In the Adversarial Self-Driving framework, a Generative Adversarial Network (GAN) is implemented to generate realistic perturbations in an image that cause a classifier CNN to misclassify data. This perturbed data is then used to train the classifier CNN further. The Adversarial Self-driving framework is applied to an image classification algorithm to improve the classification accuracy on perturbed images and is later applied to train a self-driving car to drive in a simulation. A small-scale self-driving car is also built to drive around a track and classify signs. The Adversarial Self-driving framework produces perturbed images through learning a dataset, as a result removing the need to train on significant amounts of data. Experiments demonstrate that the Adversarial Self-driving framework identifies situations where CNNs are vulnerable to perturbations and generates new examples of these situations for the CNN to train on. The additional data generated by the Adversarial Self-driving framework provides sufficient data for the CNN to generalize to the environment. Therefore, it is a viable tool to increase the resilience of CNNs to perturbations. Particularly, in the real-world self-driving car, the application of the Adversarial Self-Driving framework resulted in an 18 % increase in accuracy, and the simulated self-driving model had no collisions in 30 minutes of driving.
24.Not All Patches are What You Need: Expediting Vision Transformers via Token Reorganizations ⬇️
Vision Transformers (ViTs) take all the image patches as tokens and construct multi-head self-attention (MHSA) among them. Complete leverage of these image tokens brings redundant computations since not all the tokens are attentive in MHSA. Examples include that tokens containing semantically meaningless or distractive image backgrounds do not positively contribute to the ViT predictions. In this work, we propose to reorganize image tokens during the feed-forward process of ViT models, which is integrated into ViT during training. For each forward inference, we identify the attentive image tokens between MHSA and FFN (i.e., feed-forward network) modules, which is guided by the corresponding class token attention. Then, we reorganize image tokens by preserving attentive image tokens and fusing inattentive ones to expedite subsequent MHSA and FFN computations. To this end, our method EViT improves ViTs from two perspectives. First, under the same amount of input image tokens, our method reduces MHSA and FFN computation for efficient inference. For instance, the inference speed of DeiT-S is increased by 50% while its recognition accuracy is decreased by only 0.3% for ImageNet classification. Second, by maintaining the same computational cost, our method empowers ViTs to take more image tokens as input for recognition accuracy improvement, where the image tokens are from higher resolution images. An example is that we improve the recognition accuracy of DeiT-S by 1% for ImageNet classification at the same computational cost of a vanilla DeiT-S. Meanwhile, our method does not introduce more parameters to ViTs. Experiments on the standard benchmarks show the effectiveness of our method. The code is available at this https URL
25.Beyond Deterministic Translation for Unsupervised Domain Adaptation ⬇️
In this work we challenge the common approach of using a one-to-one mapping ('translation') between the source and target domains in unsupervised domain adaptation (UDA). Instead, we rely on stochastic translation to capture inherent translation ambiguities. This allows us to (i) train more accurate target networks by generating multiple outputs conditioned on the same source image, leveraging both accurate translation and data augmentation for appearance variability, (ii) impute robust pseudo-labels for the target data by averaging the predictions of a source network on multiple translated versions of a single target image and (iii) train and ensemble diverse networks in the target domain by modulating the degree of stochasticity in the translations. We report improvements over strong recent baselines, leading to state-of-the-art UDA results on two challenging semantic segmentation benchmarks.
26.Normalized K-Means for Noise-Insensitive Multi-Dimensional Feature Learning ⬇️
Many measurement modalities which perform imaging by probing an object pixel-by-pixel, such as via Photoacoustic Microscopy, produce a multi-dimensional feature (typically a time-domain signal) at each pixel. In principle, the many degrees of freedom in the time-domain signal would admit the possibility of significant multi-modal information being implicitly present, much more than a single scalar "brightness", regarding the underlying targets being observed. However, the measured signal is neither a weighted-sum of basis functions (such as principal components) nor one of a set of prototypes (K-means), which has motivated the novel clustering method proposed here, capable of learning centroids (signal shapes) that are related to the underlying, albeit unknown, target characteristics in a scalable and noise-robust manner.
27.Ab-initio Contrast Estimation and Denoising of Cryo-EM Images ⬇️
Background and Objective: The contrast of cryo-EM images vary from one to another, primarily due to the uneven thickness of ice layers. The variation of contrast can affect the quality of 2-D class averaging, 3-D ab-initio modeling, and 3-D heterogeneity analysis. Contrast estimation is currently performed during 3-D iterative refinement. As a result, the estimates are not available for class averaging and ab-initio modeling. However, these methods require good initial estimates of 3-D volumes and 3-D rotations of molecules. This paper aims to solve the contrast estimation problem in the ab-initio stage, without estimating the 3-D volume.
Methods: The key observation underlying our analysis is that the 2-D covariance matrix of the raw images is related to the covariance of the underlying clean images, the noise variance, and the contrast variability between images. We show that the contrast variability can be derived from the 2-D covariance matrix and use the existing Covariance Wiener Filtering (CWF) framework to estimate it. We also demonstrate a modification of CWF to estimate the contrast of individual images.
Results: Our method improves the contrast estimation by a large margin, compared to the previous CWF method. Its estimation accuracy is often comparable to that of an oracle that knows the ground truth covariance of the clean images. The more accurate contrast estimation also improves the quality of image denoising as demonstrated in both synthetic and experimental datasets.
Conclusions: This paper proposes an effective method for contrast estimation directly from noisy images without using any 3-D volume information. It enables contrast correction in the earlier stage of single particle analysis, and may improve the accuracy of downstream processing.
28.Enhancing Deformable Convolution based Video Frame Interpolation with Coarse-to-fine 3D CNN ⬇️
This paper presents a new deformable convolution-based video frame interpolation (VFI) method, using a coarse to fine 3D CNN to enhance the multi-flow prediction. This model first extracts spatio-temporal features at multiple scales using a 3D CNN, and estimates multi-flows using these features in a coarse-to-fine manner. The estimated multi-flows are then used to warp the original input frames as well as context maps, and the warped results are fused by a synthesis network to produce the final output. This VFI approach has been fully evaluated against 12 state-of-the-art VFI methods on three commonly used test databases. The results evidently show the effectiveness of the proposed method, which offers superior interpolation performance over other state of the art algorithms, with PSNR gains up to 0.19dB.
29.Don't Lie to Me! Robust and Efficient Explainability with Verified Perturbation Analysis ⬇️
A variety of methods have been proposed to try to explain how deep neural networks make their decisions. Key to those approaches is the need to sample the pixel space efficiently in order to derive importance maps. However, it has been shown that the sampling methods used to date introduce biases and other artifacts, leading to inaccurate estimates of the importance of individual pixels and severely limit the reliability of current explainability methods. Unfortunately, the alternative -- to exhaustively sample the image space is computationally prohibitive. In this paper, we introduce EVA (Explaining using Verified perturbation Analysis) -- the first explainability method guarantee to have an exhaustive exploration of a perturbation space. Specifically, we leverage the beneficial properties of verified perturbation analysis -- time efficiency, tractability and guaranteed complete coverage of a manifold -- to efficiently characterize the input variables that are most likely to drive the model decision. We evaluate the approach systematically and demonstrate state-of-the-art results on multiple benchmarks.
30.Privacy Preserving Visual Question Answering ⬇️
We introduce a novel privacy-preserving methodology for performing Visual Question Answering on the edge. Our method constructs a symbolic representation of the visual scene, using a low-complexity computer vision model that jointly predicts classes, attributes and predicates. This symbolic representation is non-differentiable, which means it cannot be used to recover the original image, thereby keeping the original image private. Our proposed hybrid solution uses a vision model which is more than 25 times smaller than the current state-of-the-art (SOTA) vision models, and 100 times smaller than end-to-end SOTA VQA models. We report detailed error analysis and discuss the trade-offs of using a distilled vision model and a symbolic representation of the visual scene.
31.Misinformation Detection in Social Media Video Posts ⬇️
With the growing adoption of short-form video by social media platforms, reducing the spread of misinformation through video posts has become a critical challenge for social media providers. In this paper, we develop methods to detect misinformation in social media posts, exploiting modalities such as video and text. Due to the lack of large-scale public data for misinformation detection in multi-modal datasets, we collect 160,000 video posts from Twitter, and leverage self-supervised learning to learn expressive representations of joint visual and textual data. In this work, we propose two new methods for detecting semantic inconsistencies within short-form social media video posts, based on contrastive learning and masked language modeling. We demonstrate that our new approaches outperform current state-of-the-art methods on both artificial data generated by random-swapping of positive samples and in the wild on a new manually-labeled test set for semantic misinformation.
32.A multi-reconstruction study of breast density estimation using Deep Learning ⬇️
Breast density estimation is one of the key tasks in recognizing individuals predisposed to breast cancer. It is often challenging because of low contrast and fluctuations in mammograms' fatty tissue background. Most of the time, the breast density is estimated manually where a radiologist assigns one of the four density categories decided by the Breast Imaging and Reporting Data Systems (BI-RADS). There have been efforts in the direction of automating a breast density classification pipeline.
Breast density estimation is one of the key tasks performed during a screening exam. Dense breasts are more susceptible to breast cancer. The density estimation is challenging because of low contrast and fluctuations in mammograms' fatty tissue background. Traditional mammograms are being replaced by tomosynthesis and its other low radiation dose variants (for example Hologic' Intelligent 2D and C-View). Because of the low-dose requirement, increasingly more screening centers are favoring the Intelligent 2D view and C-View. Deep-learning studies for breast density estimation use only a single modality for training a neural network. However, doing so restricts the number of images in the dataset. In this paper, we show that a neural network trained on all the modalities at once performs better than a neural network trained on any single modality. We discuss these results using the area under the receiver operator characteristics curves.
33.Label Propagation for Annotation-Efficient Nuclei Segmentation from Pathology Images ⬇️
Nuclei segmentation is a crucial task for whole slide image analysis in digital pathology. Generally, the segmentation performance of fully-supervised learning heavily depends on the amount and quality of the annotated data. However, it is time-consuming and expensive for professional pathologists to provide accurate pixel-level ground truth, while it is much easier to get coarse labels such as point annotations. In this paper, we propose a weakly-supervised learning method for nuclei segmentation that only requires point annotations for training. The proposed method achieves label propagation in a coarse-to-fine manner as follows. First, coarse pixel-level labels are derived from the point annotations based on the Voronoi diagram and the k-means clustering method to avoid overfitting. Second, a co-training strategy with an exponential moving average method is designed to refine the incomplete supervision of the coarse labels. Third, a self-supervised visual representation learning method is tailored for nuclei segmentation of pathology images that transforms the hematoxylin component images into the H&E stained images to gain better understanding of the relationship between the nuclei and cytoplasm. We comprehensively evaluate the proposed method using two public datasets. Both visual and quantitative results demonstrate the superiority of our method to the state-of-the-art methods, and its competitive performance compared to the fully-supervised methods. The source codes for implementing the experiments will be released after acceptance.
34.Generative modeling with projected entangled-pair states ⬇️
We argue and demonstrate that projected entangled-pair states (PEPS) outperform matrix product states significantly for the task of generative modeling of datasets with an intrinsic two-dimensional structure such as images. Our approach builds on a recently introduced algorithm for sampling PEPS, which allows for the efficient optimization and sampling of the distributions.
35.Image translation of Ultrasound to Pseudo Anatomical Display Using Artificial Intelligence ⬇️
Ultrasound is the second most used modality in medical imaging. It is cost effective, hazardless, portable and implemented routinely in numerous clinical procedures. Nonetheless, image quality is characterized by granulated appearance, poor SNR and speckle noise. Specific for malignant tumors, the margins are blurred and indistinct. Thus, there is a great need for improving ultrasound image quality. We hypothesize that this can be achieved by translation into a more realistic anatomic display, using neural networks. In order to achieve this goal, the preferable approach would be to use a set of paired images. However, this is practically impossible in our case. Therefore, CycleGAN was used, to learn each domain properties separately and enforce cross domain cycle consistency. The two datasets which were used for training the model were "Breast Ultrasound Images" (BUSI) and a set of optic images of poultry breast tissue samples acquired at our lab. The generated pseudo anatomical images provide improved visual discrimination of the lesions with clearer border definition and pronounced contrast. Furthermore, the algorithm manages to overcome the acoustic shadows artifacts commonly appearing in ultrasonic images. In order to evaluate the preservation of the anatomical features, the lesions in the ultrasonic images and the generated pseudo anatomical images were both automatically segmented and compared. This comparison yielded median dice score of 0.78 for the benign tumors and 0.43 for the malignancies. Median lesion center error of 2.38% and 8.42% for the benign and malignancies respectively and median area error index of 0.77% and 5.06% for the benign and malignancies respectively. In conclusion, these generated pseudo anatomical images, which are presented in a more intuitive way, preserve tissue anatomy and can potentially simplify the diagnosis and improve the clinical outcome.
36.Learning to Generalize across Domains on Single Test Samples ⬇️
We strive to learn a model from a set of source domains that generalizes well to unseen target domains. The main challenge in such a domain generalization scenario is the unavailability of any target domain data during training, resulting in the learned model not being explicitly adapted to the unseen target domains. We propose learning to generalize across domains on single test samples. We leverage a meta-learning paradigm to learn our model to acquire the ability of adaptation with single samples at training time so as to further adapt itself to each single test sample at test time. We formulate the adaptation to the single test sample as a variational Bayesian inference problem, which incorporates the test sample as a conditional into the generation of model parameters. The adaptation to each test sample requires only one feed-forward computation at test time without any fine-tuning or self-supervised training on additional data from the unseen domains. Extensive ablation studies demonstrate that our model learns the ability to adapt models to each single sample by mimicking domain shifts during training. Further, our model achieves at least comparable -- and often better -- performance than state-of-the-art methods on multiple benchmarks for domain generalization.
37.Diagnosing Batch Normalization in Class Incremental Learning ⬇️
Extensive researches have applied deep neural networks (DNNs) in class incremental learning (Class-IL). As building blocks of DNNs, batch normalization (BN) standardizes intermediate feature maps and has been widely validated to improve training stability and convergence. However, we claim that the direct use of standard BN in Class-IL models is harmful to both the representation learning and the classifier training, thus exacerbating catastrophic forgetting. In this paper we investigate the influence of BN on Class-IL models by illustrating such BN dilemma. We further propose BN Tricks to address the issue by training a better feature extractor while eliminating classification bias. Without inviting extra hyperparameters, we apply BN Tricks to three baseline rehearsal-based methods, ER, DER++ and iCaRL. Through comprehensive experiments conducted on benchmark datasets of Seq-CIFAR-10, Seq-CIFAR-100 and Seq-Tiny-ImageNet, we show that BN Tricks can bring significant performance gains to all adopted baselines, revealing its potential generality along this line of research.
38.ADAM Challenge: Detecting Age-related Macular Degeneration from Fundus Images ⬇️
Age-related macular degeneration (AMD) is the leading cause of visual impairment among elderly in the world. Early detection of AMD is of great importance as the vision loss caused by AMD is irreversible and permanent. Color fundus photography is the most cost-effective imaging modality to screen for retinal disorders. \textcolor{red}{Recently, some algorithms based on deep learning had been developed for fundus image analysis and automatic AMD detection. However, a comprehensive annotated dataset and a standard evaluation benchmark are still missing.} To deal with this issue, we set up the Automatic Detection challenge on Age-related Macular degeneration (ADAM) for the first time, held as a satellite event of the ISBI 2020 conference. The ADAM challenge consisted of four tasks which cover the main topics in detecting AMD from fundus images, including classification of AMD, detection and segmentation of optic disc, localization of fovea, and detection and segmentation of lesions. The ADAM challenge has released a comprehensive dataset of 1200 fundus images with the category labels of AMD, the pixel-wise segmentation masks of the full optic disc and lesions (drusen, exudate, hemorrhage, scar, and other), as well as the location coordinates of the macular fovea. A uniform evaluation framework has been built to make a fair comparison of different models. During the ADAM challenge, 610 results were submitted for online evaluation, and finally, 11 teams participated in the onsite challenge. This paper introduces the challenge, dataset, and evaluation methods, as well as summarizes the methods and analyzes the results of the participating teams of each task. In particular, we observed that ensembling strategy and clinical prior knowledge can better improve the performances of the deep learning models.
39.Meta Knowledge Distillation ⬇️
Recent studies pointed out that knowledge distillation (KD) suffers from two degradation problems, the teacher-student gap and the incompatibility with strong data augmentations, making it not applicable to training state-of-the-art models, which are trained with advanced augmentations. However, we observe that a key factor, i.e., the temperatures in the softmax functions for generating probabilities of both the teacher and student models, was mostly overlooked in previous methods. With properly tuned temperatures, such degradation problems of KD can be much mitigated. However, instead of relying on a naive grid search, which shows poor transferability, we propose Meta Knowledge Distillation (MKD) to meta-learn the distillation with learnable meta temperature parameters. The meta parameters are adaptively adjusted during training according to the gradients of the learning objective. We validate that MKD is robust to different dataset scales, different teacher/student architectures, and different types of data augmentation. With MKD, we achieve the best performance with popular ViT architectures among compared methods that use only ImageNet-1K as training data, ranging from tiny to large models. With ViT-L, we achieve 86.5% with 600 epochs of training, 0.6% better than MAE that trains for 1,650 epochs.
40.Can Deep Learning be Applied to Model-Based Multi-Object Tracking? ⬇️
Multi-object tracking (MOT) is the problem of tracking the state of an unknown and time-varying number of objects using noisy measurements, with important applications such as autonomous driving, tracking animal behavior, defense systems, and others. In recent years, deep learning (DL) has been increasingly used in MOT for improving tracking performance, but mostly in settings where the measurements are high-dimensional and there are no available models of the measurement likelihood and the object dynamics. The model-based setting instead has not attracted as much attention, and it is still unclear if DL methods can outperform traditional model-based Bayesian methods, which are the state of the art (SOTA) in this context. In this paper, we propose a Transformer-based DL tracker and evaluate its performance in the model-based setting, comparing it to SOTA model-based Bayesian methods in a variety of different tasks. Our results show that the proposed DL method can match the performance of the model-based methods in simple tasks, while outperforming them when the task gets more complicated, either due to an increase in the data association complexity, or to stronger nonlinearities of the models of the environment.
41.Cross-Modal Common Representation Learning with Triplet Loss Functions ⬇️
Common representation learning (CRL) learns a shared embedding between two or more modalities to improve in a given task over using only one of the modalities. CRL from different data types such as images and time-series data (e.g., audio or text data) requires a deep metric learning loss that minimizes the distance between the modality embeddings. In this paper, we propose to use the triplet loss, which uses positive and negative identities to create sample pairs with different labels, for CRL between image and time-series modalities. By adapting the triplet loss for CRL, higher accuracy in the main (time-series classification) task can be achieved by exploiting additional information of the auxiliary (image classification) task. Our experiments on synthetic data and handwriting recognition data from sensor-enhanced pens show an improved classification accuracy, faster convergence, and a better generalizability.
42.IPD:An Incremental Prototype based DBSCAN for large-scale data with cluster representatives ⬇️
DBSCAN is a fundamental density-based clustering technique that identifies any arbitrary shape of the clusters. However, it becomes infeasible while handling big data. On the other hand, centroid-based clustering is important for detecting patterns in a dataset since unprocessed data points can be labeled to their nearest centroid. However, it can not detect non-spherical clusters. For a large data, it is not feasible to store and compute labels of every samples. These can be done as and when the information is required. The purpose can be accomplished when clustering act as a tool to identify cluster representatives and query is served by assigning cluster labels of nearest representative. In this paper, we propose an Incremental Prototype-based DBSCAN (IPD) algorithm which is designed to identify arbitrary-shaped clusters for large-scale data. Additionally, it chooses a set of representatives for each cluster.
43.Deeply-Supervised Knowledge Distillation ⬇️
Knowledge distillation aims to enhance the performance of a lightweight student model by exploiting the knowledge from a pre-trained cumbersome teacher model. However, in the traditional knowledge distillation, teacher predictions are only used to provide the supervisory signal for the last layer of the student model, which may result in those shallow student layers lacking accurate training guidance in the layer-by-layer back propagation and thus hinders effective knowledge transfer. To address this issue, we propose Deeply-Supervised Knowledge Distillation (DSKD), which fully utilizes class predictions and feature maps of the teacher model to supervise the training of shallow student layers. A loss-based weight allocation strategy is developed in DSKD to adaptively balance the learning process of each shallow layer, so as to further improve the student performance. Extensive experiments show that the performance of DSKD consistently exceeds state-of-the-art methods on various teacher-student models, confirming the effectiveness of our proposed method.
44.Spatial Transformer K-Means ⬇️
K-means defines one of the most employed centroid-based clustering algorithms with performances tied to the data's embedding. Intricate data embeddings have been designed to push
$K$ -means performances at the cost of reduced theoretical guarantees and interpretability of the results. Instead, we propose preserving the intrinsic data space and augment K-means with a similarity measure invariant to non-rigid transformations. This enables (i) the reduction of intrinsic nuisances associated with the data, reducing the complexity of the clustering task and increasing performances and producing state-of-the-art results, (ii) clustering in the input space of the data, leading to a fully interpretable clustering algorithm, and (iii) the benefit of convergence guarantees.
45.Segmentation and Risk Score Prediction of Head and Neck Cancers in PET/CT Volumes with 3D U-Net and Cox Proportional Hazard Neural Networks ⬇️
We utilized a 3D nnU-Net model with residual layers supplemented by squeeze and excitation (SE) normalization for tumor segmentation from PET/CT images provided by the Head and Neck Tumor segmentation chal-lenge (HECKTOR). Our proposed loss function incorporates the Unified Fo-cal and Mumford-Shah losses to take the advantage of distribution, region, and boundary-based loss functions. The results of leave-one-out-center-cross-validation performed on different centers showed a segmentation performance of 0.82 average Dice score (DSC) and 3.16 median Hausdorff Distance (HD), and our results on the test set achieved 0.77 DSC and 3.01 HD. Following lesion segmentation, we proposed training a case-control proportional hazard Cox model with an MLP neural net backbone to predict the hazard risk score for each discrete lesion. This hazard risk prediction model (CoxCC) was to be trained on a number of PET/CT radiomic features extracted from the segmented lesions, patient and lesion demographics, and encoder features provided from the penultimate layer of a multi-input 2D PET/CT convolutional neural network tasked with predicting time-to-event for each lesion. A 10-fold cross-validated CoxCC model resulted in a c-index validation score of 0.89, and a c-index score of 0.61 on the HECKTOR challenge test dataset.
46.A Survey of Semen Quality Evaluation in Microscopic Videos Using Computer Assisted Sperm Analysis ⬇️
The Computer Assisted Sperm Analysis (CASA) plays a crucial role in male reproductive health diagnosis and Infertility treatment. With the development of the computer industry in recent years, a great of accurate algorithms are proposed. With the assistance of those novel algorithms, it is possible for CASA to achieve a faster and higher quality result. Since image processing is the technical basis of CASA, including pre-processing,feature extraction, target detection and tracking, these methods are important technical steps in dealing with CASA. The various works related to Computer Assisted Sperm Analysis methods in the last 30 years (since 1988) are comprehensively introduced and analysed in this survey. To facilitate understanding, the methods involved are analysed in the sequence of general steps in sperm analysis. In other words, the methods related to sperm detection (localization) are first analysed, and then the methods of sperm tracking are analysed. Beside this, we analyse and prospect the present situation and future of CASA. According to our work, the feasible for applying in sperm microscopic video of methods mentioned in this review is explained. Moreover, existing challenges of object detection and tracking in microscope video are potential to be solved inspired by this survey.
47.General-purpose, long-context autoregressive modeling with Perceiver AR ⬇️
Real-world data is high-dimensional: a book, image, or musical performance can easily contain hundreds of thousands of elements even after compression. However, the most commonly used autoregressive models, Transformers, are prohibitively expensive to scale to the number of inputs and layers needed to capture this long-range structure. We develop Perceiver AR, an autoregressive, modality-agnostic architecture which uses cross-attention to map long-range inputs to a small number of latents while also maintaining end-to-end causal masking. Perceiver AR can directly attend to over a hundred thousand tokens, enabling practical long-context density estimation without the need for hand-crafted sparsity patterns or memory mechanisms. When trained on images or music, Perceiver AR generates outputs with clear long-term coherence and structure. Our architecture also obtains state-of-the-art likelihood on long-sequence benchmarks, including 64 x 64 ImageNet images and PG-19 books.
48.Deep Learning-Assisted Co-registration of Full-Spectral Autofluorescence Lifetime Microscopic Images with H&E-Stained Histology Images ⬇️
Autofluorescence lifetime images reveal unique characteristics of endogenous fluorescence in biological samples. Comprehensive understanding and clinical diagnosis rely on co-registration with the gold standard, histology images, which is extremely challenging due to the difference of both images. Here, we show an unsupervised image-to-image translation network that significantly improves the success of the co-registration using a conventional optimisation-based regression network, applicable to autofluorescence lifetime images at different emission wavelengths. A preliminary blind comparison by experienced researchers shows the superiority of our method on co-registration. The results also indicate that the approach is applicable to various image formats, like fluorescence intensity images. With the registration, stitching outcomes illustrate the distinct differences of the spectral lifetime across an unstained tissue, enabling macro-level rapid visual identification of lung cancer and cellular-level characterisation of cell variants and common types. The approach could be effortlessly extended to lifetime images beyond this range and other staining technologies.
49.A Subjective Quality Study for Video Frame Interpolation ⬇️
Video frame interpolation (VFI) is one of the fundamental research areas in video processing and there has been extensive research on novel and enhanced interpolation algorithms. The same is not true for quality assessment of the interpolated content. In this paper, we describe a subjective quality study for VFI based on a newly developed video database, BVI-VFI. BVI-VFI contains 36 reference sequences at three different frame rates and 180 distorted videos generated using five conventional and learning based VFI algorithms. Subjective opinion scores have been collected from 60 human participants, and then employed to evaluate eight popular quality metrics, including PSNR, SSIM and LPIPS which are all commonly used for assessing VFI methods. The results indicate that none of these metrics provide acceptable correlation with the perceived quality on interpolated content, with the best-performing metric, LPIPS, offering a SROCC value below 0.6. Our findings show that there is an urgent need to develop a bespoke perceptual quality metric for VFI. The BVI-VFI dataset is publicly available and can be accessed at this https URL.