Update README.md
Browse files
README.md
CHANGED
|
@@ -1,120 +1,120 @@
|
|
| 1 |
-
---
|
| 2 |
-
license:
|
| 3 |
-
language:
|
| 4 |
-
|
| 5 |
-
tags:
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
pipeline_tag: image-to-image
|
| 12 |
-
datasets:
|
| 13 |
-
|
| 14 |
-
---
|
| 15 |
-
|
| 16 |
-
# VITON-Extends — Model weights & inference bundle
|
| 17 |
-
|
| 18 |
-
**Enhancing Pose Adaptability in Virtual Try-On Systems**
|
| 19 |
-
|
| 20 |
-
| Author | Affiliation | ORCID |
|
| 21 |
-
|--------|-------------|--------|
|
| 22 |
-
| **Nguyen Dinh Hieu** | FPT University, Hanoi, Vietnam | [0009-0002-6683-8036](https://orcid.org/0009-0002-6683-8036) |
|
| 23 |
-
| **Tran Minh Khuong** | FPT University, Hanoi, Vietnam | — |
|
| 24 |
-
| **Phan Duy Hung** | FPT University, Hanoi, Vietnam | [0000-0002-6033-6484](https://orcid.org/0000-0002-6033-6484) |
|
| 25 |
-
|
| 26 |
-
**Contact:** [hieundhe180318@fpt.edu.vn](mailto:hieundhe180318@fpt.edu.vn), [khuongtmhe180089@fpt.edu.vn](mailto:khuongtmhe180089@fpt.edu.vn), [hungpd2@fe.edu.vn](mailto:hungpd2@fe.edu.vn)
|
| 27 |
-
|
| 28 |
-
---
|
| 29 |
-
|
| 30 |
-
## What is in this Hub repository?
|
| 31 |
-
|
| 32 |
-
This **model** repo ships two top-level folders:
|
| 33 |
-
|
| 34 |
-
| Folder on Hub | Role |
|
| 35 |
-
|----------------|------|
|
| 36 |
-
| **`VITON-Extends_test/`** | Code and assets to run **inference / demo** (`test.py`, networks, options, etc.). |
|
| 37 |
-
| **`VITON-Extends-Train/`** | **Training-side** bundle (scripts, configs, or checkpoints—whatever you packaged locally under that directory). |
|
| 38 |
-
|
| 39 |
-
Download the full repo (or a subset with `allow_patterns`) and point your local paths to these folders as in the [GitHub README](https://github.com/nguyendinhhieu1309/VITON-Extends).
|
| 40 |
-
|
| 41 |
-
**Dataset (images)** is hosted separately: **[NguyenDinhHieu/VITON-Extends-DB](https://huggingface.co/datasets/NguyenDinhHieu/VITON-Extends-DB)** (see that dataset card for `Train.zip` / `Test.zip` and extraction).
|
| 42 |
-
|
| 43 |
-
---
|
| 44 |
-
|
| 45 |
-
## Abstract
|
| 46 |
-
|
| 47 |
-
Garment fitting in virtual try-on often fails under **complex poses**, **occlusions**, and **misalignment** between person and garment. VITON-Extends improves **pose adaptability** and **garment warping** with a **global appearance flow** model, **StyleGAN-style** global modulation, and a **local flow refinement** stage. On the VITON benchmark, results are strong especially in challenging poses.
|
| 48 |
-
|
| 49 |
-
**Paper:** [Springer LNCS (IUKM 2025), DOI 10.1007/978-981-96-4606-7_21](https://doi.org/10.1007/978-981-96-4606-7_21)
|
| 50 |
-
**Code:** [github.com/nguyendinhhieu1309/VITON-Extends](https://github.com/nguyendinhhieu1309/VITON-Extends)
|
| 51 |
-
|
| 52 |
-
---
|
| 53 |
-
|
| 54 |
-
## Quick download (Python)
|
| 55 |
-
|
| 56 |
-
```python
|
| 57 |
-
from huggingface_hub import snapshot_download
|
| 58 |
-
|
| 59 |
-
path = snapshot_download(
|
| 60 |
-
repo_id="NguyenDinhHieu/VITON-Extends",
|
| 61 |
-
local_dir="./VITON-Extends_hf",
|
| 62 |
-
)
|
| 63 |
-
# Then use ./VITON-Extends_hf/VITON-Extends_test/ and ./VITON-Extends_hf/VITON-Extends-Train/
|
| 64 |
-
```
|
| 65 |
-
|
| 66 |
-
---
|
| 67 |
-
|
| 68 |
-
## Environment (reference)
|
| 69 |
-
|
| 70 |
-
Versions below match the **paper / reference** setup; your local `VITON-Extends_*` trees may ship their own `requirements.txt`—prefer those for exact pins.
|
| 71 |
-
|
| 72 |
-
| Component | Reference version |
|
| 73 |
-
|-----------|-------------------|
|
| 74 |
-
| PyTorch | 2.2.1+cu118 (example) |
|
| 75 |
-
| TorchVision | 0.17.1+cu118 |
|
| 76 |
-
| CuPy | 13.3.0 |
|
| 77 |
-
| OpenCV | 4.10.0 |
|
| 78 |
-
| Python | 3.12 (or as in project env) |
|
| 79 |
-
|
| 80 |
-
---
|
| 81 |
-
|
| 82 |
-
## Training & testing (outline)
|
| 83 |
-
|
| 84 |
-
1. **Data:** Use **[VITON-Extends-DB](https://huggingface.co/datasets/NguyenDinhHieu/VITON-Extends-DB)** — unzip `Train.zip` / `Test.zip`, set `dataroot` to `train/` or `test/` as in the dataset card.
|
| 85 |
-
2. **Checkpoints:** Place warping / generation weights where the GitHub repo expects (e.g. under `checkpoints/VITON-Extends/`).
|
| 86 |
-
3. **Train:** Run the shell scripts under `scripts/` from the GitHub repository (parser-based then parser-free stages).
|
| 87 |
-
4. **Test:** From the downloaded **`VITON-Extends_test/`** tree, follow repo instructions, e.g.
|
| 88 |
-
`python test.py --name demo --resize_or_crop None --batchSize 1 --gpu_ids 0`
|
| 89 |
-
|
| 90 |
-
For **FID** and extra assets, see links in the [GitHub README](https://github.com/nguyendinhhieu1309/VITON-Extends).
|
| 91 |
-
|
| 92 |
-
---
|
| 93 |
-
|
| 94 |
-
## Results (qualitative)
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
|
| 98 |
-
---
|
| 99 |
-
|
| 100 |
-
## Citation
|
| 101 |
-
|
| 102 |
-
```bibtex
|
| 103 |
-
@inproceedings{hieu2025vitonextends,
|
| 104 |
-
title = {Enhancing Pose Adaptability in Virtual Try-On Systems},
|
| 105 |
-
author = {Hieu, Nguyen Dinh and Khuong, Tran Minh and Hung, Phan Duy},
|
| 106 |
-
booktitle = {Integrated Uncertainty in Knowledge Modelling and Decision Making (IUKM 2025)},
|
| 107 |
-
series = {Lecture Notes in Computer Science},
|
| 108 |
-
volume = {15585},
|
| 109 |
-
publisher = {Springer},
|
| 110 |
-
address = {Singapore},
|
| 111 |
-
year = {2025},
|
| 112 |
-
doi = {10.1007/978-981-96-4606-7_21}
|
| 113 |
-
}
|
| 114 |
-
```
|
| 115 |
-
|
| 116 |
-
---
|
| 117 |
-
|
| 118 |
-
## Acknowledgements
|
| 119 |
-
|
| 120 |
-
Built on virtual try-on and flow-based clothed-person generation ideas; base code lineage includes **ClothFlow**. Full credits appear in the [GitHub repository](https://github.com/nguyendinhhieu1309/VITON-Extends).
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
+
tags:
|
| 6 |
+
- pytorch
|
| 7 |
+
- virtual-try-on
|
| 8 |
+
- viton
|
| 9 |
+
- image-to-image
|
| 10 |
+
- fashion
|
| 11 |
+
pipeline_tag: image-to-image
|
| 12 |
+
datasets:
|
| 13 |
+
- NguyenDinhHieu/VITON-Extends-DB
|
| 14 |
+
---
|
| 15 |
+
|
| 16 |
+
# VITON-Extends — Model weights & inference bundle
|
| 17 |
+
|
| 18 |
+
**Enhancing Pose Adaptability in Virtual Try-On Systems**
|
| 19 |
+
|
| 20 |
+
| Author | Affiliation | ORCID |
|
| 21 |
+
|--------|-------------|--------|
|
| 22 |
+
| **Nguyen Dinh Hieu** | FPT University, Hanoi, Vietnam | [0009-0002-6683-8036](https://orcid.org/0009-0002-6683-8036) |
|
| 23 |
+
| **Tran Minh Khuong** | FPT University, Hanoi, Vietnam | — |
|
| 24 |
+
| **Phan Duy Hung** | FPT University, Hanoi, Vietnam | [0000-0002-6033-6484](https://orcid.org/0000-0002-6033-6484) |
|
| 25 |
+
|
| 26 |
+
**Contact:** [hieundhe180318@fpt.edu.vn](mailto:hieundhe180318@fpt.edu.vn), [khuongtmhe180089@fpt.edu.vn](mailto:khuongtmhe180089@fpt.edu.vn), [hungpd2@fe.edu.vn](mailto:hungpd2@fe.edu.vn)
|
| 27 |
+
|
| 28 |
+
---
|
| 29 |
+
|
| 30 |
+
## What is in this Hub repository?
|
| 31 |
+
|
| 32 |
+
This **model** repo ships two top-level folders:
|
| 33 |
+
|
| 34 |
+
| Folder on Hub | Role |
|
| 35 |
+
|----------------|------|
|
| 36 |
+
| **`VITON-Extends_test/`** | Code and assets to run **inference / demo** (`test.py`, networks, options, etc.). |
|
| 37 |
+
| **`VITON-Extends-Train/`** | **Training-side** bundle (scripts, configs, or checkpoints—whatever you packaged locally under that directory). |
|
| 38 |
+
|
| 39 |
+
Download the full repo (or a subset with `allow_patterns`) and point your local paths to these folders as in the [GitHub README](https://github.com/nguyendinhhieu1309/VITON-Extends).
|
| 40 |
+
|
| 41 |
+
**Dataset (images)** is hosted separately: **[NguyenDinhHieu/VITON-Extends-DB](https://huggingface.co/datasets/NguyenDinhHieu/VITON-Extends-DB)** (see that dataset card for `Train.zip` / `Test.zip` and extraction).
|
| 42 |
+
|
| 43 |
+
---
|
| 44 |
+
|
| 45 |
+
## Abstract
|
| 46 |
+
|
| 47 |
+
Garment fitting in virtual try-on often fails under **complex poses**, **occlusions**, and **misalignment** between person and garment. VITON-Extends improves **pose adaptability** and **garment warping** with a **global appearance flow** model, **StyleGAN-style** global modulation, and a **local flow refinement** stage. On the VITON benchmark, results are strong especially in challenging poses.
|
| 48 |
+
|
| 49 |
+
**Paper:** [Springer LNCS (IUKM 2025), DOI 10.1007/978-981-96-4606-7_21](https://doi.org/10.1007/978-981-96-4606-7_21)
|
| 50 |
+
**Code:** [github.com/nguyendinhhieu1309/VITON-Extends](https://github.com/nguyendinhhieu1309/VITON-Extends)
|
| 51 |
+
|
| 52 |
+
---
|
| 53 |
+
|
| 54 |
+
## Quick download (Python)
|
| 55 |
+
|
| 56 |
+
```python
|
| 57 |
+
from huggingface_hub import snapshot_download
|
| 58 |
+
|
| 59 |
+
path = snapshot_download(
|
| 60 |
+
repo_id="NguyenDinhHieu/VITON-Extends",
|
| 61 |
+
local_dir="./VITON-Extends_hf",
|
| 62 |
+
)
|
| 63 |
+
# Then use ./VITON-Extends_hf/VITON-Extends_test/ and ./VITON-Extends_hf/VITON-Extends-Train/
|
| 64 |
+
```
|
| 65 |
+
|
| 66 |
+
---
|
| 67 |
+
|
| 68 |
+
## Environment (reference)
|
| 69 |
+
|
| 70 |
+
Versions below match the **paper / reference** setup; your local `VITON-Extends_*` trees may ship their own `requirements.txt`—prefer those for exact pins.
|
| 71 |
+
|
| 72 |
+
| Component | Reference version |
|
| 73 |
+
|-----------|-------------------|
|
| 74 |
+
| PyTorch | 2.2.1+cu118 (example) |
|
| 75 |
+
| TorchVision | 0.17.1+cu118 |
|
| 76 |
+
| CuPy | 13.3.0 |
|
| 77 |
+
| OpenCV | 4.10.0 |
|
| 78 |
+
| Python | 3.12 (or as in project env) |
|
| 79 |
+
|
| 80 |
+
---
|
| 81 |
+
|
| 82 |
+
## Training & testing (outline)
|
| 83 |
+
|
| 84 |
+
1. **Data:** Use **[VITON-Extends-DB](https://huggingface.co/datasets/NguyenDinhHieu/VITON-Extends-DB)** — unzip `Train.zip` / `Test.zip`, set `dataroot` to `train/` or `test/` as in the dataset card.
|
| 85 |
+
2. **Checkpoints:** Place warping / generation weights where the GitHub repo expects (e.g. under `checkpoints/VITON-Extends/`).
|
| 86 |
+
3. **Train:** Run the shell scripts under `scripts/` from the GitHub repository (parser-based then parser-free stages).
|
| 87 |
+
4. **Test:** From the downloaded **`VITON-Extends_test/`** tree, follow repo instructions, e.g.
|
| 88 |
+
`python test.py --name demo --resize_or_crop None --batchSize 1 --gpu_ids 0`
|
| 89 |
+
|
| 90 |
+
For **FID** and extra assets, see links in the [GitHub README](https://github.com/nguyendinhhieu1309/VITON-Extends).
|
| 91 |
+
|
| 92 |
+
---
|
| 93 |
+
|
| 94 |
+
## Results (qualitative)
|
| 95 |
+
|
| 96 |
+
|
| 97 |
+
|
| 98 |
+
---
|
| 99 |
+
|
| 100 |
+
## Citation
|
| 101 |
+
|
| 102 |
+
```bibtex
|
| 103 |
+
@inproceedings{hieu2025vitonextends,
|
| 104 |
+
title = {Enhancing Pose Adaptability in Virtual Try-On Systems},
|
| 105 |
+
author = {Hieu, Nguyen Dinh and Khuong, Tran Minh and Hung, Phan Duy},
|
| 106 |
+
booktitle = {Integrated Uncertainty in Knowledge Modelling and Decision Making (IUKM 2025)},
|
| 107 |
+
series = {Lecture Notes in Computer Science},
|
| 108 |
+
volume = {15585},
|
| 109 |
+
publisher = {Springer},
|
| 110 |
+
address = {Singapore},
|
| 111 |
+
year = {2025},
|
| 112 |
+
doi = {10.1007/978-981-96-4606-7_21}
|
| 113 |
+
}
|
| 114 |
+
```
|
| 115 |
+
|
| 116 |
+
---
|
| 117 |
+
|
| 118 |
+
## Acknowledgements
|
| 119 |
+
|
| 120 |
+
Built on virtual try-on and flow-based clothed-person generation ideas; base code lineage includes **ClothFlow**. Full credits appear in the [GitHub repository](https://github.com/nguyendinhhieu1309/VITON-Extends).
|