Codename Error's Here!

🛍️ VITON-HD+: Virtual Try-On for Full-Body Outfits

Enhanced version of VITON-HD with support for upper clothing + pants and improved accuracy. This is the basic model of Virtual Try-on that embedded with StyloMate; Codename Error's project of Hackvidia

📋 Prerequisites

• Python 3.9+
• NVIDIA GPU (≥8GB VRAM)
• CUDA 12.8 & cuDNN 8.9.7+
• PyTorch 2.0+

🚀 Quick Start

1. Clone Repository

git clone https://github.com/yourusername/VITON-HD-Plus.git
cd VITON-HD-Plus

2. Install Dependencies

conda create -n vitonhd python=3.9
conda activate vitonhd
pip install -r requirements.txt

3. Download Pretrained Models

wget https://huggingface.co/yourusername/vitonhd-plus/resolve/main/checkpoints.zip
unzip checkpoints.zip -d ./checkpoints

4. Prepare Dataset

datasets/
├── test/
│   ├── image/          # Full-body images (e.g., 01234_00.jpg)
│   ├── cloth_upper/    # Upper clothing images
│   ├── cloth_pants/    # Pants images
│   ├── image-parse/    # Segmentation maps (include pants label)
└── test_pairs.txt      # Format: human_image.jpg upper_cloth.jpg pants.jpg

5. Run Inference

Basic

python test.py \
  --name output_demo \
  --dataset_mode full_body \
  --load_width 768 \
  --load_height 1024```

Advanced

python test.py \
  --use_pants 1 \               # Enable pants processing
  --texture_weight 0.7 \        # Fabric preservation strength
  --gmm_grid_size 5 \           # Warping precision
  --alias_resolution high       # Output quality

Advanced Configuration

python train.py \
  --name my_train \
  --semantic_nc 14 \        # 13 (original) + 1 pants class
  --dataset_dir ./custom_data \
  --save_epoch_freq 5

Training Your Own Model

Prepare training data:

python scripts/preprocess.py \
  --input_dir ./raw_data \
  --output_dir ./datasets/train

Start training:

python train.py \
  --name my_model \
  --epochs 100 \
  --batch_size 4 \
  --save_freq 10

Citation

By Seunghwan Choi, Sunghyun Park, Minsoo Lee, Jaegul Choo

@inproceedings{choi2021viton,
  title={VITON-HD: High-Resolution Virtual Try-On via Misalignment-Aware Normalization},
  author={Choi, Seunghwan and Park, Sunghyun and Lee, Minsoo and Choo, Jaegul},
  booktitle={Proc. of the IEEE conference on computer vision and pattern recognition (CVPR)},
  year={2021}
}

License

Creative Commons By-NC 4.0 © 2025 Ifan Hakim

📧 Contact For support: ifanhakm@gmail.com