feat(doc): updated FAQ

README.md

@@ -4,7 +4,7 @@
 
 <h1 align="center">Generative AI Image Toolset with GANs and Diffusion for Real-World Applications</h1>
 
-**JoliGEN** provides powerful training capabilities for generative AI image-to-image models
+**JoliGEN** provides is an integrated framework for training custom generative AI image-to-image models
 
 Main Features:
 

docs/source/FAQ.rst

@@ -1,5 +1,15 @@
-.. _qa:
+.. _faq:
 
 ############################
  Frequently Asked Questions
 ############################
+
+- **What are the training times for the examples in JoliGEN training ?**
+
+  It is reasonable to train for 10 to 15 days on 2 to 4 GPUs in 256x256 or 360x360. In 64x64 or 128x128, a couple days may suffice, always a good starting point.
+
+  In general:
+
+  - With GANs, convergence can be visually assessed within 1 or 2 days, then fine-grained details start to appear
+  - With diffusion for object insertion, training is smoother due to the straight supervision, and good results are obtained with a couple of days, then 300 to 400 epochs are reasonable
+

docs/source/index.rst

@@ -5,24 +5,23 @@
 .. image:: https://github.com/jolibrain/joliGEN/actions/workflows/github-actions-black-formatting.yml/badge.svg
    :target: https://github.com/jolibrain/joliGEN/actions/workflows/github-actions-black-formatting.yml
 
-`JoliGEN <https://github.com/jolibrain/joliGEN/>`_ provides easy-to-use
-generative AI for image to image transformations.
+`JoliGEN <https://github.com/jolibrain/joliGEN/>`_ is an integrated framework for training custom generative AI image-to-image models
 
 ***************
  Main Features
 ***************
 
--  JoliGEN support both GAN and Diffusion models for unpaired and paired
+-  JoliGEN support both **GAN and Diffusion models** for unpaired and paired
    image to image translation tasks, including domain and style
    adaptation with conservation of semantics such as image and object
    classes, masks, ...
 
 -  JoliGEN generative capabilities are targeted at real world
-   applications such as Augmented Reality, Dataset Smart Augmentation
-   and object insertion, Synthetic to real transforms.
+   applications such as **Controled Image Generation**, **Augmented Reality**, **Dataset Smart Augmentation**
+   and object insertion, **Synthetic to Real** transforms.
 
 -  JoliGEN allows for fast and stable training with astonishing results.
-   A server with REST API is provided that allows for simplified
+   A **server with REST API** is provided that allows for simplified
    deployment and usage.
 
 -  JoliGEN has a large scope of options and parameters. To not get
@@ -34,16 +33,16 @@ generative AI for image to image transformations.
  Use cases
 ***********
 
--  AR and metaverse: replace any image element with super-realistic
+-  **AR and metaverse**: replace any image element with super-realistic
    objects
--  Smart data augmentation: test / train sets augmentation
--  Image manipulation: seamlessly insert or remove objects/elements in
+-  **Smart data augmentation**: test / train sets augmentation
+-  **Image manipulation**: seamlessly insert or remove objects/elements in
    images
--  Image to image translation while preserving semantic, e.g. existing
+-  **Image to image translation** while preserving semantic, e.g. existing
    source dataset annotations
 -  Simulation to reality translation while preserving elements, metrics,
    ...
--  Image to image translation to cope with scarce data
+-  **Image generation to enrich datasets**, e.g. counter dataset imbalance, increase test sets, ...
 
 This is achieved by combining conditioned generator architectures for
 fine-grained control, bags of discriminators, configurable neural
@@ -120,9 +119,11 @@ Code is making use of `pytorch-CycleGAN-and-pix2pix
 `AttentionGAN <https://github.com/Ha0Tang/AttentionGAN>`_, `MoNCE
 <https://github.com/fnzhan/MoNCE>`_ among others.
 
-Some elements from JoliGEN are supported by the French National AI
+Elements from JoliGEN are supported by the French National AI
 program `"Confiance.AI" <https://www.confiance.ai/en/>`_
 
+Contact: [email protected]
+
 .. toctree::
    :maxdepth: 4
    :caption: Get Started

docs/source/inference.rst

@@ -32,15 +32,3 @@ Using a pretrained glasses insertion model (see above):
    python3 gen_single_image_diffusion.py --model-in-file /path/to/model/latest_net_G_A.pth --img-in /path/to/source.jpg --mask-in /path/to/mask.jpg --dir-out /path/to/target_dir/ --img-width 256 --img-height 256
 
 The mask image has 1 where to insert the object and 0 elsewhere.
-
-**************
- Export model
-**************
-
-.. code:: bash
-
-   python3 -m scripts.export_jit_model --model-in-file "/path/to/model_checkpoint.pth" --model-out-file exported_model.pt --model-type mobile_resnet_9blocks --img-size 360
-
-Then ``exported_model.pt`` can be reloaded without JoliGEN to perform
-inference with an external software, e.g. `DeepDetect
-<https://github.com/jolibrain/deepdetect>`_ with torch backend.