SO-HandNet - GitLite

KO | EN

gitlite — search

#python #go #android #deep-learning #linux #docker #macos #kubernetes #awesome-list #security #ios #mcp #openai #open-source

SO-HandNet

(★ 53)

Code repository for our paper entilted "SO-HandNet: Self-Organizing Network for 3D Hand Pose Estimation with Semi-supervised Learning", ICCV 2019.

파일 탐색기

최종 버전 다운로드 (.zip)

- augmentation.py
- ICVL_loader.py
- - ICVL_loader_v2.py
- datalist.ipynb
- ICVL_en_de.py
- ICVL_en_es.py
- ICVL_Get_test_result.py
- ICVL_semi_en_de.py
- ICVL_semi_en_es.py
- ICVL_semi_train_all.py
- ICVL_semi_train_all2.py
- ICVL_train_all.py
- ICVL_train_all2.py
- autoencoder.py
- en_estimater_debug.py
- en_estimater_semi.py
- handestimater_debug.py
- layers.py
- losses.py
- networks_debug.py
- operations.py
- - convert_depth_to_uvd.m
  - convert_jntuvd_to_jntxyz.m
  - convert_uvd_to_xyz.m
  - convert_xyz_to_uvd.m
  - evaluate_predictions.m
  - ICVL_data_split.py
  - ICVL_get_list.py
  - ICVL_test_process.m
  - ICVL_TEST_visualize_example.m
  - ICVL_train_process.m
  - ICVL_TRAIN_visualize_example.m
  - plot_hand.m
  - plot_hand_sides.m
  - plot_hand_sides_color.m
  - plot_side.m
  - plot_side_uv.m
  - plot_uv_sides.m
  - test_bug.m
- html.py
- potential_field.py
- som.py
- util.py
- visualizer.py
README.md

# 설치 가이드

1. 코드 내려받기

git clone https://github.com/TerenceCYJ/SO-HandNet

깃허브에서 프로젝트 코드 전체를 내 컴퓨터로 내려받습니다.

cd SO-HandNet

방금 내려받은 프로젝트 폴더 안으로 이동합니다.

2. 공식 설치 스크립트

쉬움 추천

사전 준비물

Python 3 pip 명령어를 쓰려면 Python이 필요합니다.

pip install --upgrade pip

PyPI에 배포된 패키지를 바로 설치합니다. 소스 클론이 필요 없습니다.

✅ 설치 후 새 터미널을 열고, 프로그램의 버전 확인 명령(예: --version)으로 정상 설치됐는지 확인하세요.

이 레포의 README에 적힌 실제 명령어를 그대로 가져왔습니다.

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# SO-HandNet
[SO-HandNet: Self-Organizing Network for 3D Hand Pose Estimation with Semi-supervised Learning](http://openaccess.thecvf.com/content_ICCV_2019/papers/Chen_SO-HandNet_Self-Organizing_Network_for_3D_Hand_Pose_Estimation_With_Semi-Supervised_ICCV_2019_paper.pdf)

ICCV 2019

## Installation
	
	conda create -n sohand python=3.6
	source activate sohand
	conda install pytorch torchvision cudatoolkit=8.0 -c pytorch
	conda install faiss-gpu -c pytorch
	pip install --upgrade pip
requirements: numba matplotlib h5py scipy dominate visdom horovod libnccl2 libnccl-dev tqdm
## Usage
### Data Preprocessing
Download ICVL dataset, and use matlab scripts to process the data (transfrom depth map into point cloud).
	
	matlab ICVL_train_process.m
	matlab ICVL_test_process.m
	
Or directly download the processed data. [Google Drive Link](https://drive.google.com/open?id=1jdEIcS6WM3v6lwirBEzU1Aw-z0TXa2xX) or [BaiduNetDesk Link](https://pan.baidu.com/s/1V8hLca_OBv5fGeR0iqhQEw)

Put data into /data as /data/ICVL/process_out/
### Train and evaluation

#### Evaluation 
pretrained models: [Google Drive Link](https://drive.google.com/open?id=1QM9U-3RH8m1Dy1-zKpVyeKuFASDa1sK_)   [BaiduNetDesk Link](https://pan.baidu.com/s/1AhDSa_G39tcEssLhABff1g)
Put data into /checkpoints
	
	python ICVL_Get_test_result.py

#### Fully-supervised Training：
	python ICVL_en_de.py
	
set "pretrain_encoder" "pretrain_decoder" as the saved model in last stage.
	
	python ICVL_en_es.py
	
set "pretrain_encoder" "pretrain_decoder" "pretrain_estimater" as the saved model in last stage.
	
	python ICVL_train_all.py
	
#### Semi-supervised training:
Change "train_label_ratio" as the ratio of labeled frames used for training, and the "trainlist" and "testlist" can be generated by "datalist.ipynb", we provide them along with the processed data.
	
	python ICVL_semi_en_de.py
	python ICVL_semi_en_es.py
	python ICVL_semi_train_all.py

## References
Here are some great resources we benefit:

[Hand PointNet](https://sites.google.com/site/geliuhaontu/home/cvpr2018)

[SO-Net](https://github.com/lijx10/SO-Net)

[PointNet](https://github.com/charlesq34/pointnet)