v2_instructions.md

Install

Create a new Conda environment:

conda create -n vtoff python=3.11
conda activate vtoff

Then, clone the repository, switch to commit used for the paper, install the required packages:

git clone https://github.com/rizavelioglu/tryoffdiff.git
cd tryoffdiff
pip install -e .

Dataset

Download the original VITON-HD dataset and extract it to "./data/vitonhd":

python tryoffdiff/dataset.py download-vitonhd  # For a different location: output-dir="<other-folder>"

As mentioned in the paper, the original dataset contains duplicates, and some training samples are leaked into the test set. Clean these with the following command:

python tryoffdiff/dataset.py clean-vitonhd  # Default: `data-dir="./data/vitonhd"`

Download the Dress Code dataset (follow steps here) and execute the following command, to match the folder structure of VITON-HD:

python tryoffdiff/dataset.py restructure_dresscode_to_vitonhd --zip-dir "/path/to/DressCode.zip"

which takes the downloaded Dress Code zip file, extracts only required folders, and restructures it to match the VITON-HD folder structure:

dresscode/
├── test/
│   ├── cloth/
│   │   ├── 048393_1.jpg
│   │   ├── 048394_1.jpg
│   │   └── ...
│   └── image/
│   │   ├── 048393_0.jpg
│   │   ├── 048394_0.jpg
│       └── ...
└── train/
    ├── cloth/
    │   ├── 000000_1.jpg
    │   ├── 000001_1.jpg
    │   └── ...
    └── image/
        ├── 000000_0.jpg
        ├── 000001_0.jpg
        └── ...

Training

For faster training, pre-extract the image features and save them instead of extracting them during training.

Step 1: Encode garment images with VAE

python tryoffdiff/dataset.py vae-encode-dataset \
 --data-dir "./data/dresscode/" \
 --model-name "sd14" \
 --batch-size 32 \
 --data-name "dresscode"

Note: We do not store SigLIP features for the Dress Code dataset, as it is 4x larger than VITON-HD and would require significantly more disk space. Instead, features are extracted on-the-fly during training.

Step 2: Train TryOffDiffv2 models with a single GPU:

To train a model for a specific garment type, such as upper body garments, use the following command:

python tryoffdiff/modeling/train.py tryoffdiffv2 \
 --save-dir "./models/" \
 --data-dir "./data/dresscode-enc-sd14/" \
 --model-class-name "TryOffDiffv2Single" \
 --mixed-precision "no" \
 --learning-rate 0.0001 \
 --train-batch-size 16 \
 --num-epochs 200 \
 --save-model-epochs 100 \
 --checkpoint-every-n-epochs 100 \
 --dataset-type "dc-upperbody"  # or "dc-lowerbody", "dc-dresses"

To train with the entire dataset, supporting all garment types, use the following command:

python tryoffdiff/modeling/train.py tryoffdiffv2 \
 --save-dir "./models/" \
 --data-dir "./data/dresscode-enc-sd14/" \
 --model-class-name "TryOffDiffv2Multi" \  # Multi-garment TryOffDiff
 --mixed-precision "no" \
 --learning-rate 0.0001 \
 --train-batch-size 16 \
 --num-epochs 200 \
 --save-model-epochs 100 \
 --checkpoint-every-n-epochs 100 \
 --dataset-type "dresscode"  # Entire dataset with all garment types

Inference

To run inference with a trained model (regardless of its model class, i.e. TryOffDiffv2Single or TryOffDiffv2Multi), simply use the following command:

python tryoffdiff/modeling/predict.py tryoffdiffv2 \
  --model-dir "/model_20250710_123456/" \
  --model-filename "model_epoch_200.pth" \
  --batch-size 8  \
  --num-inference-steps 20 \
  --seed 42 \
  --guidance-scale 2.5

which runs inference on the entire test set and saves predictions to "<model-dir>/preds/" as .png files.

Evaluation

Evaluate the predictions using:

python tryoffdiff/modeling/eval.py \
 --gt-dir "./data/dresscode/test/cloth/" \
 --pred-dir "<prediction-dir>" \
 --batch-size 32 \
 --num-workers 4

which prints the results to the console. Specifically, we use the following libraries for the implementations of the metrics presented in the paper:

• pyiqa: SSIM, MS-SSIM, CW-SSIM, and LPIPS,
• clean-fid: FID, CLIP-FID, and KID,
• DISTS-pytorch: DISTS

In addition, we offer a simple GUI for visualizing predictions alongside their evaluation metrics. This tool displays the ground truth and predicted images side-by-side while providing metrics for the entire test set:

python tryoffdiff/modeling/eval_vis.py \
 --gt-dir "./data/dresscode/test/cloth/" \
 --pred-dir "<prediction-dir>"

---

Back

To main page