1.Ranking Info Noise Contrastive Estimation: Boosting Contrastive Learning via Ranked Positives ⬇️
This paper introduces Ranking Info Noise Contrastive Estimation (RINCE), a new member in the family of InfoNCE losses that preserves a ranked ordering of positive samples. In contrast to the standard InfoNCE loss, which requires a strict binary separation of the training pairs into similar and dissimilar samples, RINCE can exploit information about a similarity ranking for learning a corresponding embedding space. We show that the proposed loss function learns favorable embeddings compared to the standard InfoNCE whenever at least noisy ranking information can be obtained or when the definition of positives and negatives is blurry. We demonstrate this for a supervised classification task with additional superclass labels and noisy similarity scores. Furthermore, we show that RINCE can also be applied to unsupervised training with experiments on unsupervised representation learning from videos. In particular, the embedding yields higher classification accuracy, retrieval rates and performs better in out-of-distribution detection than the standard InfoNCE loss.
2.Constrained Structure Learning for Scene Graph Generation ⬇️
As a structured prediction task, scene graph generation aims to build a visually-grounded scene graph to explicitly model objects and their relationships in an input image. Currently, the mean field variational Bayesian framework is the de facto methodology used by the existing methods, in which the unconstrained inference step is often implemented by a message passing neural network. However, such formulation fails to explore other inference strategies, and largely ignores the more general constrained optimization models. In this paper, we present a constrained structure learning method, for which an explicit constrained variational inference objective is proposed. Instead of applying the ubiquitous message-passing strategy, a generic constrained optimization method - entropic mirror descent - is utilized to solve the constrained variational inference step. We validate the proposed generic model on various popular scene graph generation benchmarks and show that it outperforms the state-of-the-art methods.
3.Vision Checklist: Towards Testable Error Analysis of Image Models to Help System Designers Interrogate Model Capabilities ⬇️
Using large pre-trained models for image recognition tasks is becoming increasingly common owing to the well acknowledged success of recent models like vision transformers and other CNN-based models like VGG and Resnet. The high accuracy of these models on benchmark tasks has translated into their practical use across many domains including safety-critical applications like autonomous driving and medical diagnostics. Despite their widespread use, image models have been shown to be fragile to changes in the operating environment, bringing their robustness into question. There is an urgent need for methods that systematically characterise and quantify the capabilities of these models to help designers understand and provide guarantees about their safety and robustness. In this paper, we propose Vision Checklist, a framework aimed at interrogating the capabilities of a model in order to produce a report that can be used by a system designer for robustness evaluations. This framework proposes a set of perturbation operations that can be applied on the underlying data to generate test samples of different types. The perturbations reflect potential changes in operating environments, and interrogate various properties ranging from the strictly quantitative to more qualitative. Our framework is evaluated on multiple datasets like Tinyimagenet, CIFAR10, CIFAR100 and Camelyon17 and for models like ViT and Resnet. Our Vision Checklist proposes a specific set of evaluations that can be integrated into the previously proposed concept of a model card. Robustness evaluations like our checklist will be crucial in future safety evaluations of visual perception modules, and be useful for a wide range of stakeholders including designers, deployers, and regulators involved in the certification of these systems. Source code of Vision Checklist would be open for public use.
4.Team Yao at Factify 2022: Utilizing Pre-trained Models and Co-attention Networks for Multi-Modal Fact Verification ⬇️
In recent years, social media has enabled users to get exposed to a myriad of misinformation and disinformation; thus, misinformation has attracted a great deal of attention in research fields and as a social issue. To address the problem, we propose a framework, Pre-CoFact, composed of two pre-trained models for extracting features from text and images, and multiple co-attention networks for fusing the same modality but different sources and different modalities. Besides, we adopt the ensemble method by using different pre-trained models in Pre-CoFact to achieve better performance. We further illustrate the effectiveness from the ablation study and examine different pre-trained models for comparison. Our team, Yao, won the fifth prize (F1-score: 74.585%) in the Factify challenge hosted by De-Factify @ AAAI 2022, which demonstrates that our model achieved competitive performance without using auxiliary tasks or extra information. The source code of our work is publicly available at this https URL
5.Deep Video Prior for Video Consistency and Propagation ⬇️
Applying an image processing algorithm independently to each video frame often leads to temporal inconsistency in the resulting video. To address this issue, we present a novel and general approach for blind video temporal consistency. Our method is only trained on a pair of original and processed videos directly instead of a large dataset. Unlike most previous methods that enforce temporal consistency with optical flow, we show that temporal consistency can be achieved by training a convolutional neural network on a video with Deep Video Prior (DVP). Moreover, a carefully designed iteratively reweighted training strategy is proposed to address the challenging multimodal inconsistency problem. We demonstrate the effectiveness of our approach on 7 computer vision tasks on videos. Extensive quantitative and perceptual experiments show that our approach obtains superior performance than state-of-the-art methods on blind video temporal consistency. We further extend DVP to video propagation and demonstrate its effectiveness in propagating three different types of information (color, artistic style, and object segmentation). A progressive propagation strategy with pseudo labels is also proposed to enhance DVP's performance on video propagation. Our source codes are publicly available at this https URL.
6.Domain generalization in deep learning-based mass detection in mammography: A large-scale multi-center study ⬇️
Computer-aided detection systems based on deep learning have shown great potential in breast cancer detection. However, the lack of domain generalization of artificial neural networks is an important obstacle to their deployment in changing clinical environments. In this work, we explore the domain generalization of deep learning methods for mass detection in digital mammography and analyze in-depth the sources of domain shift in a large-scale multi-center setting. To this end, we compare the performance of eight state-of-the-art detection methods, including Transformer-based models, trained in a single domain and tested in five unseen domains. Moreover, a single-source mass detection training pipeline is designed to improve the domain generalization without requiring images from the new domain. The results show that our workflow generalizes better than state-of-the-art transfer learning-based approaches in four out of five domains while reducing the domain shift caused by the different acquisition protocols and scanner manufacturers. Subsequently, an extensive analysis is performed to identify the covariate shifts with bigger effects on the detection performance, such as due to differences in patient age, breast density, mass size, and mass malignancy. Ultimately, this comprehensive study provides key insights and best practices for future research on domain generalization in deep learning-based breast cancer detection.
7.A Probabilistic Framework for Dynamic Object Recognition in 3D Environment With A Novel Continuous Ground Estimation Method ⬇️
In this thesis a probabilistic framework is developed and proposed for Dynamic Object Recognition in 3D Environments. A software package is developed using C++ and Python in ROS that performs the detection and tracking task. Furthermore, a novel Gaussian Process Regression (GPR) based method is developed to detect ground points in different urban scenarios of regular, sloped and rough. The ground surface behavior is assumed to only demonstrate local input-dependent smoothness. kernel's length-scales are obtained. Bayesian inference is implemented sing \textit{Maximum a Posteriori} criterion. The log-marginal likelihood function is assumed to be a multi-task objective function, to represent a whole-frame unbiased view of the ground at each frame because adjacent segments may not have similar ground structure in an uneven scene while having shared hyper-parameter values. Simulation results shows the effectiveness of the proposed method in uneven and rough scenes which outperforms similar Gaussian process based ground segmentation methods.
8.ASOC: Adaptive Self-aware Object Co-localization ⬇️
The primary goal of this paper is to localize objects in a group of semantically similar images jointly, also known as the object co-localization problem. Most related existing works are essentially weakly-supervised, relying prominently on the neighboring images' weak-supervision. Although weak supervision is beneficial, it is not entirely reliable, for the results are quite sensitive to the neighboring images considered. In this paper, we combine it with a self-awareness phenomenon to mitigate this issue. By self-awareness here, we refer to the solution derived from the image itself in the form of saliency cue, which can also be unreliable if applied alone. Nevertheless, combining these two paradigms together can lead to a better co-localization ability. Specifically, we introduce a dynamic mediator that adaptively strikes a proper balance between the two static solutions to provide an optimal solution. Therefore, we call this method \textit{ASOC}: Adaptive Self-aware Object Co-localization. We perform exhaustive experiments on several benchmark datasets and validate that weak-supervision supplemented with self-awareness has superior performance outperforming several compared competing methods.
9.Beyond ImageNet Attack: Towards Crafting Adversarial Examples for Black-box Domains ⬇️
Adversarial examples have posed a severe threat to deep neural networks due to their transferable nature. Currently, various works have paid great efforts to enhance the cross-model transferability, which mostly assume the substitute model is trained in the same domain as the target model. However, in reality, the relevant information of the deployed model is unlikely to leak. Hence, it is vital to build a more practical black-box threat model to overcome this limitation and evaluate the vulnerability of deployed models. In this paper, with only the knowledge of the ImageNet domain, we propose a Beyond ImageNet Attack (BIA) to investigate the transferability towards black-box domains (unknown classification tasks). Specifically, we leverage a generative model to learn the adversarial function for disrupting low-level features of input images. Based on this framework, we further propose two variants to narrow the gap between the source and target domains from the data and model perspectives, respectively. Extensive experiments on coarse-grained and fine-grained domains demonstrate the effectiveness of our proposed methods. Notably, our methods outperform state-of-the-art approaches by up to 7.71% (towards coarse-grained domains) and 25.91% (towards fine-grained domains) on average. Our code is available at \url{this https URL}.
10.ResiDualGAN: Resize-Residual DualGAN for Cross-Domain Remote Sensing Images Semantic Segmentation ⬇️
The performance of a semantic segmentation model for remote sensing (RS) images pretrained on an annotated dataset would greatly decrease when testing on another unannotated dataset because of the domain gap. Adversarial generative methods, e.g., DualGAN, are utilized for unpaired image-to-image translation to minimize the pixel-level domain gap, which is one of the common approaches for unsupervised domain adaptation (UDA). However, existing image translation methods are facing two problems when performing RS images translation: 1) ignoring the scale discrepancy between two RS datasets which greatly affect the accuracy performance of scale-invariant objects, 2) ignoring the characteristic of real-to-real translation of RS images which brings an unstable factor for the training of the models. In this paper, ResiDualGAN is proposed for RS images translation, where a resizer module is used for addressing the scale discrepancy of RS datasets, and a residual connection is used for strengthening the stability of real-to-real images translation and improving the performance in cross-domain semantic segmentation tasks. Combining with an output space adaptation method, the proposed method greatly improves the accuracy performance on common benchmarks, which demonstrates the superiority and reliability of ResiDuanGAN. At the end of the paper, a thorough discussion is also conducted to give a reasonable explanation for the improvement of ResiDualGAN.
11.Anomaly Detection in Retinal Images using Multi-Scale Deep Feature Sparse Coding ⬇️
Convolutional Neural Network models have successfully detected retinal illness from optical coherence tomography (OCT) and fundus images. These CNN models frequently rely on vast amounts of labeled data for training, difficult to obtain, especially for rare diseases. Furthermore, a deep learning system trained on a data set with only one or a few diseases cannot detect other diseases, limiting the system's practical use in disease identification. We have introduced an unsupervised approach for detecting anomalies in retinal images to overcome this issue. We have proposed a simple, memory efficient, easy to train method which followed a multi-step training technique that incorporated autoencoder training and Multi-Scale Deep Feature Sparse Coding (MDFSC), an extended version of normal sparse coding, to accommodate diverse types of retinal datasets. We achieve relative AUC score improvement of 7.8%, 6.7% and 12.1% over state-of-the-art SPADE on Eye-Q, IDRiD and OCTID datasets respectively.
12.Head and eye egocentric gesture recognition for human-robot interaction using eyewear cameras ⬇️
Non-verbal communication plays a particularly important role in a wide range of scenarios in Human-Robot Interaction (HRI). Accordingly, this work addresses the problem of human gesture recognition. In particular, we focus on head and eye gestures, and adopt an egocentric (first-person) perspective using eyewear cameras. We argue that this egocentric view offers a number of conceptual and technical benefits over scene- or robot-centric perspectives.
A motion-based recognition approach is proposed, which operates at two temporal granularities. Locally, frame-to-frame homographies are estimated with a convolutional neural network (CNN). The output of this CNN is input to a long short-term memory (LSTM) to capture longer-term temporal visual relationships, which are relevant to characterize gestures.
Regarding the configuration of the network architecture, one particularly interesting finding is that using the output of an internal layer of the homography CNN increases the recognition rate with respect to using the homography matrix itself. While this work focuses on action recognition, and no robot or user study has been conducted yet, the system has been de signed to meet real-time constraints. The encouraging results suggest that the proposed egocentric perspective is viable, and this proof-of-concept work provides novel and useful contributions to the exciting area of HRI.
13.Eye-focused Detection of Bell's Palsy in Videos ⬇️
In this paper, we present how Bell's Palsy, a neurological disorder, can be detected just from a subject's eyes in a video. We notice that Bell's Palsy patients often struggle to blink their eyes on the affected side. As a result, we can observe a clear contrast between the blinking patterns of the two eyes. Although previous works did utilize images/videos to detect this disorder, none have explicitly focused on the eyes. Most of them require the entire face. One obvious advantage of having an eye-focused detection system is that subjects' anonymity is not at risk. Also, our AI decisions based on simple blinking patterns make them explainable and straightforward. Specifically, we develop a novel feature called blink similarity, which measures the similarity between the two blinking patterns. Our extensive experiments demonstrate that the proposed feature is quite robust, for it helps in Bell's Palsy detection even with very few labels. Our proposed eye-focused detection system is not only cheaper but also more convenient than several existing methods.
14.RelTR: Relation Transformer for Scene Graph Generation ⬇️
Different objects in the same scene are more or less related to each other, but only a limited number of these relationships are noteworthy. Inspired by DETR, which excels in object detection, we view scene graph generation as a set prediction problem and propose an end-to-end scene graph generation model RelTR which has an encoder-decoder architecture. The encoder reasons about the visual feature context while the decoder infers a fixed-size set of triplets subject-predicate-object using different types of attention mechanisms with coupled subject and object queries. We design a set prediction loss performing the matching between the ground truth and predicted triplets for the end-to-end training. In contrast to most existing scene graph generation methods, RelTR is a one-stage method that predicts a set of relationships directly only using visual appearance without combining entities and labeling all possible predicates. Extensive experiments on the Visual Genome and Open Images V6 datasets demonstrate the superior performance and fast inference of our model.
15.In Defense of Kalman Filtering for Polyp Tracking from Colonoscopy Videos ⬇️
Real-time and robust automatic detection of polyps from colonoscopy videos are essential tasks to help improve the performance of doctors during this exam. The current focus of the field is on the development of accurate but inefficient detectors that will not enable a real-time application. We advocate that the field should instead focus on the development of simple and efficient detectors that an be combined with effective trackers to allow the implementation of real-time polyp detectors. In this paper, we propose a Kalman filtering tracker that can work together with powerful, but efficient detectors, enabling the implementation of real-time polyp detectors. In particular, we show that the combination of our Kalman filtering with the detector PP-YOLO shows state-of-the-art (SOTA) detection accuracy and real-time processing. More specifically, our approach has SOTA results on the CVC-ClinicDB dataset, with a recall of 0.740, precision of 0.869,
$F_1$ score of 0.799, an average precision (AP) of 0.837, and can run in real time (i.e., 30 frames per second). We also evaluate our method on a subset of the Hyper-Kvasir annotated by our clinical collaborators, resulting in SOTA results, with a recall of 0.956, precision of 0.875,$F_1$ score of 0.914, AP of 0.952, and can run in real time.
16.An Analysis on Ensemble Learning optimized Medical Image Classification with Deep Convolutional Neural Networks ⬇️
Novel and high-performance medical image classification pipelines are heavily utilizing ensemble learning strategies. The idea of ensemble learning is to assemble diverse models or multiple predictions and, thus, boost prediction performance. However, it is still an open question to what extent as well as which ensemble learning strategies are beneficial in deep learning based medical image classification pipelines. In this work, we proposed a reproducible medical image classification pipeline for analyzing the performance impact of the following ensemble learning techniques: Augmenting, Stacking, and Bagging. The pipeline consists of state-of-the-art preprocessing and image augmentation methods as well as 9 deep convolution neural network architectures. It was applied on four popular medical imaging datasets with varying complexity. Furthermore, 12 pooling functions for combining multiple predictions were analyzed, ranging from simple statistical functions like unweighted averaging up to more complex learning-based functions like support vector machines. Our results revealed that Stacking achieved the largest performance gain of up to 13% F1-score increase. Augmenting showed consistent improvement capabilities by up to 4% and is also applicable to single model based pipelines. Cross-validation based Bagging demonstrated to be the most complex ensemble learning method, which resulted in an F1-score decrease in all analyzed datasets (up to -10%). Furthermore, we demonstrated that simple statistical pooling functions are equal or often even better than more complex pooling functions. We concluded that the integration of Stacking and Augmentation ensemble learning techniques is a powerful method for any medical image classification pipeline to improve robustness and boost performance.
17.DocSegTr: An Instance-Level End-to-End Document Image Segmentation Transformer ⬇️
Understanding documents with rich layouts is an essential step towards information extraction. Business intelligence processes often require the extraction of useful semantic content from documents at a large scale for subsequent decision-making tasks. In this context, instance-level segmentation of different document objects(title, sections, figures, tables and so on) has emerged as an interesting problem for the document layout analysis community. To advance the research in this direction, we present a transformer-based model for end-to-end segmentation of complex layouts in document images. To our knowledge, this is the first work on transformer-based document segmentation. Extensive experimentation on the PubLayNet dataset shows that our model achieved comparable or better segmentation performance than the existing state-of-the-art approaches. We hope our simple and flexible framework could serve as a promising baseline for instance-level recognition tasks in document images.
18.Non-linear Motion Estimation for Video Frame Interpolation using Space-time Convolutions ⬇️
Video frame interpolation aims to synthesize one or multiple frames between two consecutive frames in a video. It has a wide range of applications including slow-motion video generation, frame-rate up-scaling and developing video codecs. Some older works tackled this problem by assuming per-pixel linear motion between video frames. However, objects often follow a non-linear motion pattern in the real domain and some recent methods attempt to model per-pixel motion by non-linear models (e.g., quadratic). A quadratic model can also be inaccurate, especially in the case of motion discontinuities over time (i.e. sudden jerks) and occlusions, where some of the flow information may be invalid or inaccurate.
In our paper, we propose to approximate the per-pixel motion using a space-time convolution network that is able to adaptively select the motion model to be used. Specifically, we are able to softly switch between a linear and a quadratic model. Towards this end, we use an end-to-end 3D CNN encoder-decoder architecture over bidirectional optical flows and occlusion maps to estimate the non-linear motion model of each pixel. Further, a motion refinement module is employed to refine the non-linear motion and the interpolated frames are estimated by a simple warping of the neighboring frames with the estimated per-pixel motion. Through a set of comprehensive experiments, we validate the effectiveness of our model and show that our method outperforms state-of-the-art algorithms on four datasets (Vimeo, DAVIS, HD and GoPro).
19.Generalised Image Outpainting with U-Transformer ⬇️
While most present image outpainting conducts horizontal extrapolation, we study the generalised image outpainting problem that extrapolates visual context all-side around a given image. To this end, we develop a novel transformer-based generative adversarial network called U-Transformer able to extend image borders with plausible structure and details even for complicated scenery images. Specifically, we design a generator as an encoder-to-decoder structure embedded with the popular Swin Transformer blocks. As such, our novel framework can better cope with image long-range dependencies which are crucially important for generalised image outpainting. We propose additionally a U-shaped structure and multi-view Temporal Spatial Predictor network to reinforce image self-reconstruction as well as unknown-part prediction smoothly and realistically. We experimentally demonstrate that our proposed method could produce visually appealing results for generalized image outpainting against the state-of-the-art image outpainting approaches.
20.Contrastive Embedding Distribution Refinement and Entropy-Aware Attention for 3D Point Cloud Classification ⬇️
Learning a powerful representation from point clouds is a fundamental and challenging problem in the field of computer vision. Different from images where RGB pixels are stored in the regular grid, for point clouds, the underlying semantic and structural information of point clouds is the spatial layout of the points. Moreover, the properties of challenging in-context and background noise pose more challenges to point cloud analysis. One assumption is that the poor performance of the classification model can be attributed to the indistinguishable embedding feature that impedes the search for the optimal classifier. This work offers a new strategy for learning powerful representations via a contrastive learning approach that can be embedded into any point cloud classification network. First, we propose a supervised contrastive classification method to implement embedding feature distribution refinement by improving the intra-class compactness and inter-class separability. Second, to solve the confusion problem caused by small inter-class compactness and inter-class separability. Second, to solve the confusion problem caused by small inter-class variations between some similar-looking categories, we propose a confusion-prone class mining strategy to alleviate the confusion effect. Finally, considering that outliers of the sample clusters in the embedding space may cause performance degradation, we design an entropy-aware attention module with information entropy theory to identify the outlier cases and the unstable samples by measuring the uncertainty of predicted probability. The results of extensive experiments demonstrate that our method outperforms the state-of-the-art approaches by achieving 82.9% accuracy on the real-world ScanObjectNN dataset and substantial performance gains up to 2.9% in DCGNN, 3.1% in PointNet++, and 2.4% in GBNet.
21.Deep Confidence Guided Distance for 3D Partial Shape Registration ⬇️
We present a novel non-iterative learnable method for partial-to-partial 3D shape registration. The partial alignment task is extremely complex, as it jointly tries to match between points and identify which points do not appear in the corresponding shape, causing the solution to be non-unique and ill-posed in most cases.
Until now, two principal methodologies have been suggested to solve this problem: sample a subset of points that are likely to have correspondences or perform soft alignment between the point clouds and try to avoid a match to an occluded part. These heuristics work when the partiality is mild or when the transformation is small but fails for severe occlusions or when outliers are present. We present a unique approach named Confidence Guided Distance Network (CGD-net), where we fuse learnable similarity between point embeddings and spatial distance between point clouds, inducing an optimized solution for the overlapping points while ignoring parts that only appear in one of the shapes. The point feature generation is done by a self-supervised architecture that repels far points to have different embeddings, therefore succeeds to align partial views of shapes, even with excessive internal symmetries or acute rotations. We compare our network to recently presented learning-based and axiomatic methods and report a fundamental boost in performance.
22.Effective Shortcut Technique for GAN ⬇️
In recent years, generative adversarial network (GAN)-based image generation techniques design their generators by stacking up multiple residual blocks. The residual block generally contains a shortcut, \ie skip connection, which effectively supports information propagation in the network. In this paper, we propose a novel shortcut method, called the gated shortcut, which not only embraces the strength point of the residual block but also further boosts the GAN performance. More specifically, based on the gating mechanism, the proposed method leads the residual block to keep (or remove) information that is relevant (or irrelevant) to the image being generated. To demonstrate that the proposed method brings significant improvements in the GAN performance, this paper provides extensive experimental results on the various standard datasets such as CIFAR-10, CIFAR-100, LSUN, and tiny-ImageNet. Quantitative evaluations show that the gated shortcut achieves the impressive GAN performance in terms of Frechet inception distance (FID) and Inception score (IS). For instance, the proposed method improves the FID and IS scores on the tiny-ImageNet dataset from 35.13 to 27.90 and 20.23 to 23.42, respectively.
23.Exploring Global Diversity and Local Context for Video Summarization ⬇️
Video summarization aims to automatically generate a diverse and concise summary which is useful in large-scale video processing. Most of methods tend to adopt self attention mechanism across video frames, which fails to model the diversity of video frames. To alleviate this problem, we revisit the pairwise similarity measurement in self attention mechanism and find that the existing inner-product affinity leads to discriminative features rather than diversified features. In light of this phenomenon, we propose global diverse attention by using the squared Euclidean distance instead to compute the affinities. Moreover, we model the local contextual information by proposing local contextual attention to remove the redundancy in the video. By combining these two attention mechanism, a video \textbf{SUM}marization model with Diversified Contextual Attention scheme is developed and named as SUM-DCA. Extensive experiments are conducted on benchmark data sets to verify the effectiveness and the superiority of SUM-DCA in terms of F-score and rank-based evaluation without any bells and whistles.
24.Dynamic Rectification Knowledge Distillation ⬇️
Knowledge Distillation is a technique which aims to utilize dark knowledge to compress and transfer information from a vast, well-trained neural network (teacher model) to a smaller, less capable neural network (student model) with improved inference efficiency. This approach of distilling knowledge has gained popularity as a result of the prohibitively complicated nature of such cumbersome models for deployment on edge computing devices. Generally, the teacher models used to teach smaller student models are cumbersome in nature and expensive to train. To eliminate the necessity for a cumbersome teacher model completely, we propose a simple yet effective knowledge distillation framework that we termed Dynamic Rectification Knowledge Distillation (DR-KD). Our method transforms the student into its own teacher, and if the self-teacher makes wrong predictions while distilling information, the error is rectified prior to the knowledge being distilled. Specifically, the teacher targets are dynamically tweaked by the agency of ground-truth while distilling the knowledge gained from traditional training. Our proposed DR-KD performs remarkably well in the absence of a sophisticated cumbersome teacher model and achieves comparable performance to existing state-of-the-art teacher-free knowledge distillation frameworks when implemented by a low-cost dynamic mannered teacher. Our approach is all-encompassing and can be utilized for any deep neural network training that requires categorization or object recognition. DR-KD enhances the test accuracy on Tiny ImageNet by 2.65% over prominent baseline models, which is significantly better than any other knowledge distillation approach while requiring no additional training costs.
25.Transformer Module Networks for Systematic Generalization in Visual Question Answering ⬇️
Transformer-based models achieve great performance on Visual Question Answering (VQA). However, when we evaluate them on systematic generalization, i.e., handling novel combinations of known concepts, their performance degrades. Neural Module Networks (NMNs) are a promising approach for systematic generalization that consists on composing modules, i.e., neural networks that tackle a sub-task. Inspired by Transformers and NMNs, we propose Transformer Module Network (TMN), a novel Transformer-based model for VQA that dynamically composes modules into a question-specific Transformer network. TMNs achieve state-of-the-art systematic generalization performance in three VQA datasets, namely, CLEVR-CoGenT, CLOSURE and GQA-SGL, in some cases improving more than 30% over standard Transformers.
26.Dissecting the impact of different loss functions with gradient surgery ⬇️
Pair-wise loss is an approach to metric learning that learns a semantic embedding by optimizing a loss function that encourages images from the same semantic class to be mapped closer than images from different classes. The literature reports a large and growing set of variations of the pair-wise loss strategies. Here we decompose the gradient of these loss functions into components that relate to how they push the relative feature positions of the anchor-positive and anchor-negative pairs. This decomposition allows the unification of a large collection of current pair-wise loss functions. Additionally, explicitly constructing pair-wise gradient updates to separate out these effects gives insights into which have the biggest impact, and leads to a simple algorithm that beats the state of the art for image retrieval on the CAR, CUB and Stanford Online products datasets.
27.Efficient divide-and-conquer registration of UAV and ground LiDAR point clouds through canopy shape context ⬇️
Registration of unmanned aerial vehicle laser scanning (ULS) and ground light detection and ranging (LiDAR) point clouds in forests is critical to create a detailed representation of a forest structure and an accurate inversion of forest parameters. However, forest occlusion poses challenges for marker-based registration methods, and some marker-free automated registration methods have low efficiency due to the process of object (e.g., tree, crown) segmentation. Therefore, we use a divide-and-conquer strategy and propose an automated and efficient method to register ULS and ground LiDAR point clouds in forests. Registration involves coarse alignment and fine registration, where the coarse alignment of point clouds is divided into vertical and horizontal alignment. The vertical alignment is achieved by ground alignment, which is achieved by the transformation relationship between normal vectors of the ground point cloud and the horizontal plane, and the horizontal alignment is achieved by canopy projection image matching. During image matching, vegetation points are first distinguished by the ground filtering algorithm, and then, vegetation points are projected onto the horizontal plane to obtain two binary images. To match the two images, a matching strategy is used based on canopy shape context features, which are described by a two-point congruent set and canopy overlap. Finally, we implement coarse alignment of ULS and ground LiDAR datasets by combining the results of ground alignment and image matching and finish fine registration. Also, the effectiveness, accuracy, and efficiency of the proposed method are demonstrated by field measurements of forest plots. Experimental results show that the ULS and ground LiDAR data in different plots are registered, of which the horizontal alignment errors are less than 0.02 m, and the average runtime of the proposed method is less than 1 second.
28.Interactive 3D Character Modeling from 2D Orthogonal Drawings with Annotations ⬇️
We propose an interactive 3D character modeling approach from orthographic drawings (e.g., front and side views) based on 2D-space annotations. First, the system builds partial correspondences between the input drawings and generates a base mesh with sweeping splines according to edge information in 2D images. Next, users annotates the desired parts on the input drawings (e.g., the eyes and mouth) by using two type of strokes, called addition and erosion, and the system re-optimizes the shape of the base mesh. By repeating the 2D-space operations (i.e., revising and modifying the annotations), users can design a desired character model. To validate the efficiency and quality of our system, we verified the generated results with state-of-the-art methods.
29.Revisiting RCAN: Improved Training for Image Super-Resolution ⬇️
Image super-resolution (SR) is a fast-moving field with novel architectures attracting the spotlight. However, most SR models were optimized with dated training strategies. In this work, we revisit the popular RCAN model and examine the effect of different training options in SR. Surprisingly (or perhaps as expected), we show that RCAN can outperform or match nearly all the CNN-based SR architectures published after RCAN on standard benchmarks with a proper training strategy and minimal architecture change. Besides, although RCAN is a very large SR architecture with more than four hundred convolutional layers, we draw a notable conclusion that underfitting is still the main problem restricting the model capability instead of overfitting. We observe supportive evidence that increasing training iterations clearly improves the model performance while applying regularization techniques generally degrades the predictions. We denote our simply revised RCAN as RCAN-it and recommend practitioners to use it as baselines for future research. Code is publicly available at this https URL.
30.Continuous Examination by Automatic Quiz Assessment Using Spiral Codes and Image Processing ⬇️
We describe a technical solution implemented at Halmstad University to automatise assessment and reporting of results of paper-based quiz exams. Paper quizzes are affordable and within reach of campus education in classrooms. Offering and taking them is accepted as they cause fewer issues with reliability and democratic access, e.g. a large number of students can take them without a trusted mobile device, internet, or battery. By contrast, correction of the quiz is a considerable obstacle. We suggest mitigating the issue by a novel image processing technique using harmonic spirals that aligns answer sheets in sub-pixel accuracy to read student identity and answers and to email results within minutes, all fully automatically. Using the described method, we carry out regular weekly examinations in two master courses at the mentioned centre without a significant workload increase. The employed solution also enables us to assign a unique identifier to each quiz (e.g. week 1, week 2. . . ) while allowing us to have an individualised quiz for each student.
31.Challenges and Opportunities for Machine Learning Classification of Behavior and Mental State from Images ⬇️
Computer Vision (CV) classifiers which distinguish and detect nonverbal social human behavior and mental state can aid digital diagnostics and therapeutics for psychiatry and the behavioral sciences. While CV classifiers for traditional and structured classification tasks can be developed with standard machine learning pipelines for supervised learning consisting of data labeling, preprocessing, and training a convolutional neural network, there are several pain points which arise when attempting this process for behavioral phenotyping. Here, we discuss the challenges and corresponding opportunities in this space, including handling heterogeneous data, avoiding biased models, labeling massive and repetitive data sets, working with ambiguous or compound class labels, managing privacy concerns, creating appropriate representations, and personalizing models. We discuss current state-of-the-art research endeavors in CV such as data curation, data augmentation, crowdsourced labeling, active learning, reinforcement learning, generative models, representation learning, federated learning, and meta-learning. We highlight at least some of the machine learning advancements needed for imaging classifiers to detect human social cues successfully and reliably.
32.ReforesTree: A Dataset for Estimating Tropical Forest Carbon Stock with Deep Learning and Aerial Imagery ⬇️
Forest biomass is a key influence for future climate, and the world urgently needs highly scalable financing schemes, such as carbon offsetting certifications, to protect and restore forests. Current manual forest carbon stock inventory methods of measuring single trees by hand are time, labour, and cost-intensive and have been shown to be subjective. They can lead to substantial overestimation of the carbon stock and ultimately distrust in forest financing. The potential for impact and scale of leveraging advancements in machine learning and remote sensing technologies is promising but needs to be of high quality in order to replace the current forest stock protocols for certifications.
In this paper, we present ReforesTree, a benchmark dataset of forest carbon stock in six agro-forestry carbon offsetting sites in Ecuador. Furthermore, we show that a deep learning-based end-to-end model using individual tree detection from low cost RGB-only drone imagery is accurately estimating forest carbon stock within official carbon offsetting certification standards. Additionally, our baseline CNN model outperforms state-of-the-art satellite-based forest biomass and carbon stock estimates for this type of small-scale, tropical agro-forestry sites. We present this dataset to encourage machine learning research in this area to increase accountability and transparency of monitoring, verification and reporting (MVR) in carbon offsetting projects, as well as scaling global reforestation financing through accurate remote sensing.
33.DIREG3D: DIrectly REGress 3D Hands from Multiple Cameras ⬇️
In this paper, we present DIREG3D, a holistic framework for 3D Hand Tracking. The proposed framework is capable of utilizing camera intrinsic parameters, 3D geometry, intermediate 2D cues, and visual information to regress parameters for accurately representing a Hand Mesh model. Our experiments show that information like the size of the 2D hand, its distance from the optical center, and radial distortion is useful for deriving highly reliable 3D poses in camera space from just monocular information. Furthermore, we extend these results to a multi-view camera setup by fusing features from different viewpoints.
34.PRNU Based Source Camera Identification for Webcam and Smartphone Videos ⬇️
This communication is about an application of image forensics where we use camera sensor fingerprints to identify source camera (SCI: Source Camera Identification) in webcam/smartphone videos. Sensor or camera fingerprints are based on computing the intrinsic noise that is always present in this kind of sensors due to manufacturing imperfections. This is an unavoidable characteristic that links each sensor with its noise pattern. PRNU (Photo Response Non-Uniformity) has become the default technique to compute a camera fingerprint. There are many applications nowadays dealing with PRNU patterns for camera identification using still images. In this work we focus on video, first on webcam video and afterwards on smartphone video. Webcams and smartphones are the most used video cameras nowadays. Three possible methods for SCI are implemented and assessed in this work.
35.IGLUE: A Benchmark for Transfer Learning across Modalities, Tasks, and Languages ⬇️
Reliable evaluation benchmarks designed for replicability and comprehensiveness have driven progress in machine learning. Due to the lack of a multilingual benchmark, however, vision-and-language research has mostly focused on English language tasks. To fill this gap, we introduce the Image-Grounded Language Understanding Evaluation benchmark. IGLUE brings together - by both aggregating pre-existing datasets and creating new ones - visual question answering, cross-modal retrieval, grounded reasoning, and grounded entailment tasks across 20 diverse languages. Our benchmark enables the evaluation of multilingual multimodal models for transfer learning, not only in a zero-shot setting, but also in newly defined few-shot learning setups. Based on the evaluation of the available state-of-the-art models, we find that translate-test transfer is superior to zero-shot transfer and that few-shot learning is hard to harness for many tasks. Moreover, downstream performance is partially explained by the amount of available unlabelled textual data for pretraining, and only weakly by the typological distance of target-source languages. We hope to encourage future research efforts in this area by releasing the benchmark to the community.
36.A Systematic Study of Bias Amplification ⬇️
Recent research suggests that predictions made by machine-learning models can amplify biases present in the training data. When a model amplifies bias, it makes certain predictions at a higher rate for some groups than expected based on training-data statistics. Mitigating such bias amplification requires a deep understanding of the mechanics in modern machine learning that give rise to that amplification. We perform the first systematic, controlled study into when and how bias amplification occurs. To enable this study, we design a simple image-classification problem in which we can tightly control (synthetic) biases. Our study of this problem reveals that the strength of bias amplification is correlated to measures such as model accuracy, model capacity, model overconfidence, and amount of training data. We also find that bias amplification can vary greatly during training. Finally, we find that bias amplification may depend on the difficulty of the classification task relative to the difficulty of recognizing group membership: bias amplification appears to occur primarily when it is easier to recognize group membership than class membership. Our results suggest best practices for training machine-learning models that we hope will help pave the way for the development of better mitigation strategies.
37.Matched Illumination ⬇️
In previous work, it was shown that a camera can theoretically be made more colorimetric - its RGBs become more linearly related to XYZ tristimuli - by placing a specially designed color filter in the optical path. While the prior art demonstrated the principle, the optimal color-correction filters were not actually manufactured. In this paper, we provide a novel way of creating the color filtering effect without making a physical filter: we modulate the spectrum of the light source by using a spectrally tunable lighting system to recast the prefiltering effect from a lighting perspective. According to our method, if we wish to measure color under a D65 light, we relight the scene with a modulated D65 spectrum where the light modulation mimics the effect of color prefiltering in the prior art. We call our optimally modulated light, the matched illumination. In the experiments, using synthetic and real measurements, we show that color measurement errors can be reduced by about 50% or more on simulated data and 25% or more on real images when the matched illumination is used.
38.DropNAS: Grouped Operation Dropout for Differentiable Architecture Search ⬇️
Neural architecture search (NAS) has shown encouraging results in automating the architecture design. Recently, DARTS relaxes the search process with a differentiable formulation that leverages weight-sharing and SGD where all candidate operations are trained simultaneously. Our empirical results show that such procedure results in the co-adaption problem and Matthew Effect: operations with fewer parameters would be trained maturely earlier. This causes two problems: firstly, the operations with more parameters may never have the chance to express the desired function since those with less have already done the job; secondly, the system will punish those underperforming operations by lowering their architecture parameter, and they will get smaller loss gradients, which causes the Matthew Effect. In this paper, we systematically study these problems and propose a novel grouped operation dropout algorithm named DropNAS to fix the problems with DARTS. Extensive experiments demonstrate that DropNAS solves the above issues and achieves promising performance. Specifically, DropNAS achieves 2.26% test error on CIFAR-10, 16.39% on CIFAR-100 and 23.4% on ImageNet (with the same training hyperparameters as DARTS for a fair comparison). It is also observed that DropNAS is robust across variants of the DARTS search space. Code is available at this https URL.
39.Unsupervised Change Detection using DRE-CUSUM ⬇️
This paper presents DRE-CUSUM, an unsupervised density-ratio estimation (DRE) based approach to determine statistical changes in time-series data when no knowledge of the pre-and post-change distributions are available. The core idea behind the proposed approach is to split the time-series at an arbitrary point and estimate the ratio of densities of distribution (using a parametric model such as a neural network) before and after the split point. The DRE-CUSUM change detection statistic is then derived from the cumulative sum (CUSUM) of the logarithm of the estimated density ratio. We present a theoretical justification as well as accuracy guarantees which show that the proposed statistic can reliably detect statistical changes, irrespective of the split point. While there have been prior works on using density ratio based methods for change detection, to the best of our knowledge, this is the first unsupervised change detection approach with a theoretical justification and accuracy guarantees. The simplicity of the proposed framework makes it readily applicable in various practical settings (including high-dimensional time-series data); we also discuss generalizations for online change detection. We experimentally show the superiority of DRE-CUSUM using both synthetic and real-world datasets over existing state-of-the-art unsupervised algorithms (such as Bayesian online change detection, its variants as well as several other heuristic methods).
40.Automatic Classification of Neuromuscular Diseases in Children Using Photoacoustic Imaging ⬇️
Neuromuscular diseases (NMDs) cause a significant burden for both healthcare systems and society. They can lead to severe progressive muscle weakness, muscle degeneration, contracture, deformity and progressive disability. The NMDs evaluated in this study often manifest in early childhood. As subtypes of disease, e.g. Duchenne Muscular Dystropy (DMD) and Spinal Muscular Atrophy (SMA), are difficult to differentiate at the beginning and worsen quickly, fast and reliable differential diagnosis is crucial. Photoacoustic and ultrasound imaging has shown great potential to visualize and quantify the extent of different diseases. The addition of automatic classification of such image data could further improve standard diagnostic procedures. We compare deep learning-based 2-class and 3-class classifiers based on VGG16 for differentiating healthy from diseased muscular tissue. This work shows promising results with high accuracies above 0.86 for the 3-class problem and can be used as a proof of concept for future approaches for earlier diagnosis and therapeutic monitoring of NMDs.
41.Density-Aware Hyper-Graph Neural Networks for Graph-based Semi-supervised Node Classification ⬇️
Graph-based semi-supervised learning, which can exploit the connectivity relationship between labeled and unlabeled data, has been shown to outperform the state-of-the-art in many artificial intelligence applications. One of the most challenging problems for graph-based semi-supervised node classification is how to use the implicit information among various data to improve the performance of classifying. Traditional studies on graph-based semi-supervised learning have focused on the pairwise connections among data. However, the data correlation in real applications could be beyond pairwise and more complicated. The density information has been demonstrated to be an important clue, but it is rarely explored in depth among existing graph-based semi-supervised node classification methods. To develop a flexible and effective model for graph-based semi-supervised node classification, we propose a novel Density-Aware Hyper-Graph Neural Networks (DA-HGNN). In our proposed approach, hyper-graph is provided to explore the high-order semantic correlation among data, and a density-aware hyper-graph attention network is presented to explore the high-order connection relationship. Extensive experiments are conducted in various benchmark datasets, and the results demonstrate the effectiveness of the proposed approach.
42.Pan-Tumor CAnine cuTaneous Cancer Histology (CATCH) Dataset ⬇️
Due to morphological similarities, the differentiation of histologic sections of cutaneous tumors into individual subtypes can be challenging. Recently, deep learning-based approaches have proven their potential for supporting pathologists in this regard. However, many of these supervised algorithms require a large amount of annotated data for robust development. We present a publicly available dataset consisting of 350 whole slide images of seven different canine cutaneous tumors complemented by 12,424 polygon annotations for 13 histologic classes including seven cutaneous tumor subtypes. Regarding sample size and annotation extent, this exceeds most publicly available datasets which are oftentimes limited to the tumor area or merely provide patch-level annotations. We validated our model for tissue segmentation, achieving a class-averaged Jaccard coefficient of 0.7047, and 0.9044 for tumor in particular. For tumor subtype classification, we achieve a slide-level accuracy of 0.9857. Since canine cutaneous tumors possess various histologic homologies to human tumors, we believe that the added value of this dataset is not limited to veterinary pathology but extends to more general fields of application.
43.Multi-Frame Quality Enhancement On Compressed Video Using Quantised Data of Deep Belief Networks ⬇️
In the age of streaming and surveillance compressed video enhancement has become a problem in need of constant improvement. Here, we investigate a way of improving the Multi-Frame Quality Enhancement approach. This approach consists of making use of the frames that have the peak quality in the region to improve those that have a lower quality in that region. This approach consists of obtaining quantized data from the videos using a deep belief network. The quantized data is then fed into the MF-CNN architecture to improve the compressed video. We further investigate the impact of using a Bi-LSTM for detecting the peak quality frames. Our approach obtains better results than the first approach of the MFQE which uses an SVM for PQF detection. On the other hand, our MFQE approach does not outperform the latest version of the MQFE approach that uses a Bi-LSTM for PQF detection.
44.Few-shot Transfer Learning for Holographic Image Reconstruction using a Recurrent Neural Network ⬇️
Deep learning-based methods in computational microscopy have been shown to be powerful but in general face some challenges due to limited generalization to new types of samples and requirements for large and diverse training data. Here, we demonstrate a few-shot transfer learning method that helps a holographic image reconstruction deep neural network rapidly generalize to new types of samples using small datasets. We pre-trained a convolutional recurrent neural network on a large dataset with diverse types of samples, which serves as the backbone model. By fixing the recurrent blocks and transferring the rest of the convolutional blocks of the pre-trained model, we reduced the number of trainable parameters by ~90% compared with standard transfer learning, while achieving equivalent generalization. We validated the effectiveness of this approach by successfully generalizing to new types of samples using small holographic datasets for training, and achieved (i) ~2.5-fold convergence speed acceleration, (ii) ~20% computation time reduction per epoch, and (iii) improved reconstruction performance over baseline network models trained from scratch. This few-shot transfer learning approach can potentially be applied in other microscopic imaging methods, helping to generalize to new types of samples without the need for extensive training time and data.
45.Controlling Directions Orthogonal to a Classifier ⬇️
We propose to identify directions invariant to a given classifier so that these directions can be controlled in tasks such as style transfer. While orthogonal decomposition is directly identifiable when the given classifier is linear, we formally define a notion of orthogonality in the non-linear case. We also provide a surprisingly simple method for constructing the orthogonal classifier (a classifier utilizing directions other than those of the given classifier). Empirically, we present three use cases where controlling orthogonal variation is important: style transfer, domain adaptation, and fairness. The orthogonal classifier enables desired style transfer when domains vary in multiple aspects, improves domain adaptation with label shifts and mitigates the unfairness as a predictor. The code is available at this http URL
46.HistoKT: Cross Knowledge Transfer in Computational Pathology ⬇️
The lack of well-annotated datasets in computational pathology (CPath) obstructs the application of deep learning techniques for classifying medical images. %Since pathologist time is expensive, dataset curation is intrinsically difficult. Many CPath workflows involve transferring learned knowledge between various image domains through transfer learning. Currently, most transfer learning research follows a model-centric approach, tuning network parameters to improve transfer results over few datasets. In this paper, we take a data-centric approach to the transfer learning problem and examine the existence of generalizable knowledge between histopathological datasets. First, we create a standardization workflow for aggregating existing histopathological data. We then measure inter-domain knowledge by training ResNet18 models across multiple histopathological datasets, and cross-transferring between them to determine the quantity and quality of innate shared knowledge. Additionally, we use weight distillation to share knowledge between models without additional training. We find that hard to learn, multi-class datasets benefit most from pretraining, and a two stage learning framework incorporating a large source domain such as ImageNet allows for better utilization of smaller datasets. Furthermore, we find that weight distillation enables models trained on purely histopathological features to outperform models using external natural image data.