feat: add DDPM tutorial on the VITON-HD dataset

docs/source/index.rst

@@ -205,3 +205,10 @@ Contact: contact@jolibrain.com
    tips
    FAQ
 
+.. toctree::
+   :maxdepth: 2
+   :caption: Tutorials
+   :glob:
+   :hidden:
+
+   tutorial_viton

docs/source/tutorial_viton.rst

@@ -0,0 +1,123 @@
+####################################
+ DDPM model on the VITON-HD dataset
+####################################
+
+*****************
+ Install joliGEN
+*****************
+
+.. code:: bash
+
+   cd
+   git clone https://github.com/jolibrain/joliGEN.git
+   cd joliGEN
+   pip install -r requirements.txt --upgrade
+
+More details: :doc:`install`
+
+*********************
+ Prepare the Dataset
+*********************
+
+Preprocess the dataset provided by `VITON-HD
+<https://github.com/shadow2496/VITON-HD#dataset>`_ (4.4 Gb) to a joliGEN
+compatible format:
+
+.. code:: bash
+
+   cd
+   wget --continue https://www.dropbox.com/s/10bfat0kg4si1bu/zalando-hd-resized.zip
+   python3 ~/joliGEN/scripts/preprocess_viton.py --zip-file zalando-hd-resized.zip --target-dir ~/datasets/VITON-HD/ --dilate 5
+
+This will produce two folders ``trainA`` and ``testA`` under the
+``~/VITON-HD/datasets`` folder.
+
+Each of these folder contains:
+
+-  ``imgs``: the original images
+-  ``mask``: the masks for the top clothes area (orange part of the
+   VITON-HD segmentation)
+-  ``paths.txt``: the pairs image/mask used for training/testing
+
+****************************
+ Train your Diffusion Model
+****************************
+
+.. code:: bash
+
+   cd ~/joliGEN
+   python3 train.py \
+        --dataroot ~/datasets/VITON-HD/ \
+        --checkpoints_dir ~/checkpoints/ \
+        --name VITON-HD \
+        --gpu_ids 0 \
+        --model_type palette \
+        --train_batch_size 8 \
+        --data_num_threads 16 \
+        --train_iter_size 1 \
+        --model_input_nc 3 \
+        --model_output_nc 3 \
+        --data_relative_paths \
+        --train_G_ema \
+        --train_optim adamw \
+        --data_dataset_mode self_supervised_labeled_mask \
+        --data_load_size 256 \
+        --data_crop_size 256 \
+        --G_netG unet_mha \
+        --data_online_creation_rand_mask_A \
+        --train_G_lr 0.0001 \
+        --train_n_epochs 100 \
+        --dataaug_no_rotate \
+        --dataaug_no_flip \
+        --output_display_freq 20000 \
+        --output_print_freq 500 \
+        --output_display_visdom_autostart
+
+-  If you have multiple GPUs, you can use them by adjusting the
+   ``--gpu_ids`` option.
+-  If you run out of memory, you should lower the ``--train_batch_size``
+   option.
+
+************************
+ Training Visualization
+************************
+
+Open http://localhost:8097/env/VITON-HD to monitor your training.
+
+.. image:: _static/tutorial_viton_visdom.png
+
+The columns contain in this order:
+
+-  original image
+-  conditioning image (unused in this experience)
+-  initial noise
+-  mask
+-  generated image
+
+More details: :doc:`quickstart_ddpm`
+
+***********
+ Inference
+***********
+
+.. code:: bash
+
+   mkdir -p ~/inferences
+   cd ~/joliGEN/scripts
+   python3 gen_single_image_diffusion.py \
+        --model-in-file ~/checkpoints/VITON-HD/latest_net_G_A.pth \
+        --img-in ~/datasets/VITON-HD/testA/imgs/00006_00.jpg \
+        --mask-in ~/datasets/VITON-HD/testA/mask/00006_00.png \
+        --dir-out ~/inferences \
+        --nb_samples 4 \
+        --img-width 256 \
+        --img-height 256
+
+This will produce 4 samples in the ``~/inferences`` folder.
+
+In the example below, original image and mask followed by 4 generated
+images:
+
+.. image:: _static/tutorial_viton_inference.png
+
+More details: :doc:`inference`