Learning to Disambiguate Strongly Interacting Hands via Probabilistic Per-Pixel Part Segmentation [3DV 2021 Oral]

Related tags

Deep Learningdigit-interacting
Overview

Learning to Disambiguate Strongly Interacting Hands via Probabilistic Per-Pixel Part Segmentation [3DV 2021 Oral]

report report

Learning to Disambiguate Strongly Interacting Hands via Probabilistic Per-Pixel Part Segmentation,
Zicong Fan, Adrian Spurr, Muhammed Kocabas, Siyu Tang, Michael J. Black, Otmar Hilliges International Conference on 3D Vision (3DV), 2021

Image

Features

DIGIT estimates the 3D poses of two interacting hands from a single RGB image. This repo provides the training, evaluation, and demo code for the project in PyTorch Lightning.

Updates

  • November 25 2021: Initial repo with training and evaluation on PyTorch Lightning 0.9.

Setting up environment

DIGIT has been implemented and tested on Ubuntu 18.04 with python >= 3.7, PyTorch Lightning 0.9 and PyTorch 1.6.

Clone the repo:

git clone https://github.com/zc-alexfan/digit-interacting

Create folders needed:

make folders

Install conda environment:

conda create -n digit python=3.7
conda deactivate
conda activate digit
conda install pytorch==1.6.0 torchvision==0.7.0 cudatoolkit=10.1 -c pytorch
pip install -r requirements.txt

Downloading InterHand2.6M

  • Download the 5fps.v1 of InterHand2.6M, following the instructions here
  • Place annotations, images, and rootnet_output from InterHand2.6M under ./data/InterHand/*:
./data/InterHand
├── annotations
├── images
│   ├── test
│   ├── train
│   └── val
├── rootnet_output
│   ├── rootnet_interhand2.6m_output_all_test.json
│   └── rootnet_interhand2.6m_output_machine_annot_val.json
|-- annotations
|-- images
|   |-- test
|   |-- train
|   `-- val
`-- rootnet_output
    |-- rootnet_interhand2.6m_output_test.json
    `-- rootnet_interhand2.6m_output_val.json
  • The folder ./data/InterHand/annotations should look like this:
./data/InterHand/annotations
|-- skeleton.txt
|-- subject.txt
|-- test
|   |-- InterHand2.6M_test_MANO_NeuralAnnot.json
|   |-- InterHand2.6M_test_camera.json
|   |-- InterHand2.6M_test_data.json
|   `-- InterHand2.6M_test_joint_3d.json
|-- train
|   |-- InterHand2.6M_train_MANO_NeuralAnnot.json
|   |-- InterHand2.6M_train_camera.json
|   |-- InterHand2.6M_train_data.json
|   `-- InterHand2.6M_train_joint_3d.json
`-- val
    |-- InterHand2.6M_val_MANO_NeuralAnnot.json
    |-- InterHand2.6M_val_camera.json
    |-- InterHand2.6M_val_data.json
    `-- InterHand2.6M_val_joint_3d.json

Preparing data and backbone for training

Download the ImageNet-pretrained backbone from here and place it under:

./saved_models/pytorch/imagenet/hrnet_w32-36af842e.pt

Package images into lmdb:

cd scripts
python package_images_lmdb.py

Preprocess annotation:

python preprocess_annot.py

Render part segmentation masks:

  • Following the README.md of render_mano_ih to prepare an LMDB of part segmentation. For question in preparing the segmentation masks, please keep issues in there.

Place the LMDB from the images, the segmentation masks, and meta_dict_*.pkl to ./data/InterHand and it should look like the structure below. The cache files meta_dict_*.pkl are by-products of the step above.

|-- annotations
|   |-- skeleton.txt
|   |-- subject.txt
|   |-- test
|   |   |-- InterHand2.6M_test_MANO_NeuralAnnot.json
|   |   |-- InterHand2.6M_test_camera.json
|   |   |-- InterHand2.6M_test_data.json
|   |   |-- InterHand2.6M_test_data.pkl
|   |   `-- InterHand2.6M_test_joint_3d.json
|   |-- train
|   |   |-- InterHand2.6M_train_MANO_NeuralAnnot.json
|   |   |-- InterHand2.6M_train_camera.json
|   |   |-- InterHand2.6M_train_data.json
|   |   |-- InterHand2.6M_train_data.pkl
|   |   `-- InterHand2.6M_train_joint_3d.json
|   `-- val
|       |-- InterHand2.6M_val_MANO_NeuralAnnot.json
|       |-- InterHand2.6M_val_camera.json
|       |-- InterHand2.6M_val_data.json
|       |-- InterHand2.6M_val_data.pkl
|       `-- InterHand2.6M_val_joint_3d.json
|-- cache
|   |-- meta_dict_test.pkl
|   |-- meta_dict_train.pkl
|   `-- meta_dict_val.pkl
|-- images
|   |-- test
|   |-- train
|   `-- val
|-- rootnet_output
|   |-- rootnet_interhand2.6m_output_test.json
|   `-- rootnet_interhand2.6m_output_val.json
`-- segm_32.lmdb

Training and evaluating

To train DIGIT, run the command below. The script runs at a batch size of 64 using accumulated gradient where each iteration is on a batch size 32:

python train.py --iter_batch 32 --batch_size 64 --gpu_ids 0 --trainsplit train --precision 16 --eval_every_epoch 2 --lr_dec_epoch 40 --max_epoch 50 --min_epoch 50

OR if you just want to do a sanity check you can run:

python train.py --iter_batch 32 --batch_size 64 --gpu_ids 0 --trainsplit minitrain --valsplit minival --precision 16 --eval_every_epoch 1 --max_epoch 50 --min_epoch 50

Each time you run train.py, it will create a new experiment under logs and each experiment is assigned a key.

Supposed your experiment key is 2e8c5136b, you can evaluate the last epoch of the model on the test set by:

python test.py --eval_on minitest --load_ckpt logs/2e8c5136b/model_dump/last.ckpt

OR

python test.py --eval_on test --load_ckpt logs/2e8c5136b/model_dump/last.ckpt

The former only does the evaluation 1000 images for a sanity check.

Similarly, you can evaluate on the validation set:

python test.py --eval_on val --load_ckpt logs/2e8c5136b/model_dump/last.ckpt

Visualizing and evaluating pre-trained DIGIT

Here we provide instructions to show qualitative results of DIGIT.

Download pre-trained DIGIT:

wget https://dataset.ait.ethz.ch/downloads/dE6qPPePCV/db7cba8c1.pt
mv db7cba8c1.pt saved_models

Visualize results:

CUDA_VISIBLE_DEVICES=0 python demo.py --eval_on minival --load_from saved_models/db7cba8c1.pt  --num_workers 0

Evaluate pre-trained digit:

CUDA_VISIBLE_DEVICES=0 python test.py --eval_on test --load_from saved_models/db7cba8c1.pt --precision 16
CUDA_VISIBLE_DEVICES=0 python test.py --eval_on val --load_from saved_models/db7cba8c1.pt --precision 16

You should have the same results as in here.

The results will be dumped to ./visualization.

Citation

@inProceedings{fan2021digit,
  title={Learning to Disambiguate Strongly Interacting Hands via Probabilistic Per-pixel Part Segmentation},
  author={Fan, Zicong and Spurr, Adrian and Kocabas, Muhammed and Tang, Siyu and Black, Michael and Hilliges, Otmar},
  booktitle={International Conference on 3D Vision (3DV)},
  year={2021}
}

License

Since our code is developed based on InterHand2.6M, which is CC-BY-NC 4.0 licensed, the same LICENSE is applied to DIGIT.

DIGIT is CC-BY-NC 4.0 licensed, as found in the LICENSE file.

References

Some code in our repo uses snippets of the following repo:

Please consider citing them if you find our code useful:

@inproceedings{Moon_2020_ECCV_InterHand2.6M,  
author = {Moon, Gyeongsik and Yu, Shoou-I and Wen, He and Shiratori, Takaaki and Lee, Kyoung Mu},  
title = {InterHand2.6M: A Dataset and Baseline for 3D Interacting Hand Pose Estimation from a Single RGB Image},  
booktitle = {European Conference on Computer Vision (ECCV)},  
year = {2020}  
}  

@inproceedings{sun2019deep,
  title={Deep High-Resolution Representation Learning for Human Pose Estimation},
  author={Sun, Ke and Xiao, Bin and Liu, Dong and Wang, Jingdong},
  booktitle={CVPR},
  year={2019}
}

@inproceedings{xiao2018simple,
    author={Xiao, Bin and Wu, Haiping and Wei, Yichen},
    title={Simple Baselines for Human Pose Estimation and Tracking},
    booktitle = {European Conference on Computer Vision (ECCV)},
    year = {2018}
}

@misc{Charles2013,
  author = {milesial},
  title = {Pytorch-UNet},
  year = {2021},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/milesial/Pytorch-UNet}}
}

Contact

For any question, you can contact [email protected].

Owner
Zicong Fan
A Ph.D. student at ETH Zurich.
Zicong Fan
GitHub Repository
Offical implementation of Shunted Self-Attention via Multi-Scale Token Aggregation

Shunted Transformer This is the offical implementation of Shunted Self-Attention via Multi-Scale Token Aggregation by Sucheng Ren, Daquan Zhou, Shengf

156 Dec 27, 2022
Code for models used in Bashiri et al., "A Flow-based latent state generative model of neural population responses to natural images".

A Flow-based latent state generative model of neural population responses to natural images Code for "A Flow-based latent state generative model of ne

Sinz Lab 5 Aug 26, 2022
Doods2 - API for detecting objects in images and video streams using Tensorflow

DOODS2 - Return of DOODS Dedicated Open Object Detection Service - Yes, it's a b

Zach 101 Jan 04, 2023
Contrastive Loss Gradient Attack (CLGA)

Contrastive Loss Gradient Attack (CLGA) Official implementation of Unsupervised Graph Poisoning Attack via Contrastive Loss Back-propagation, WWW22 Bu

12 Dec 23, 2022
Generative Art Using Neural Visual Grammars and Dual Encoders

Generative Art Using Neural Visual Grammars and Dual Encoders Arnheim 1 The original algorithm from the paper Generative Art Using Neural Visual Gramm

DeepMind 231 Jan 05, 2023
Deep Reinforcement Learning based autonomous navigation for quadcopters using PPO algorithm.

PPO-based Autonomous Navigation for Quadcopters This repository contains an implementation of Proximal Policy Optimization (PPO) for autonomous naviga

Bilal Kabas 16 Nov 11, 2022
A solution to the 2D Ising model of ferromagnetism, implemented using the Metropolis algorithm

Solving the Ising model on a 2D lattice using the Metropolis Algorithm Introduction The Ising model is a simplified model of ferromagnetism, the pheno

Rohit Prabhu 5 Nov 13, 2022
Animatable Neural Radiance Fields for Modeling Dynamic Human Bodies

To make the comparison with Animatable NeRF easier on the Human3.6M dataset, we save the quantitative results at here, which also contains the results of other methods, including Neural Body, D-NeRF,

ZJU3DV 359 Jan 08, 2023
【Arxiv】Exploring Separable Attention for Multi-Contrast MR Image Super-Resolution

SANet Exploring Separable Attention for Multi-Contrast MR Image Super-Resolution Dependencies numpy==1.18.5 scikit_image==0.16.2 torchvision==0.8.1 to

36 Jan 05, 2023
A Confidence-based Iterative Solver of Depths and Surface Normals for Deep Multi-view Stereo

idn-solver Paper | Project Page This repository contains the code release of our ICCV 2021 paper: A Confidence-based Iterative Solver of Depths and Su

zhaowang 43 Nov 17, 2022
Implementation of "A MLP-like Architecture for Dense Prediction"

A MLP-like Architecture for Dense Prediction (arXiv) Updates (22/07/2021) Initial release. Model Zoo We provide CycleMLP models pretrained on ImageNet

Shoufa Chen 244 Dec 27, 2022
[ICLR 2022] Pretraining Text Encoders with Adversarial Mixture of Training Signal Generators

AMOS This repository contains the scripts for fine-tuning AMOS pretrained models on GLUE and SQuAD 2.0 benchmarks. Paper: Pretraining Text Encoders wi

Microsoft 22 Sep 15, 2022
This respository includes implementations on Manifoldron: Direct Space Partition via Manifold Discovery

Manifoldron: Direct Space Partition via Manifold Discovery This respository includes implementations on Manifoldron: Direct Space Partition via Manifo

dayang_wang 4 Apr 28, 2022
Topic Modelling for Humans

gensim – Topic Modelling in Python Gensim is a Python library for topic modelling, document indexing and similarity retrieval with large corpora. Targ

RARE Technologies 13.8k Jan 03, 2023
The versatile ocean simulator, in pure Python, powered by JAX.

Veros is the versatile ocean simulator -- it aims to be a powerful tool that makes high-performance ocean modeling approachable and fun. Because Veros

TeamOcean 245 Dec 20, 2022
Trans-Encoder: Unsupervised sentence-pair modelling through self- and mutual-distillations

Trans-Encoder: Unsupervised sentence-pair modelling through self- and mutual-distillations Code repo for paper Trans-Encoder: Unsupervised sentence-pa

Amazon 101 Dec 29, 2022
Artificial Intelligence search algorithm base on Pacman

Pacman Search Artificial Intelligence search algorithm base on Pacman Source The Pacman Projects by the University of California, Berkeley. Layouts Di

Day Fundora 6 Nov 17, 2022
Chess reinforcement learning by AlphaGo Zero methods.

About Chess reinforcement learning by AlphaGo Zero methods. This project is based on these main resources: DeepMind's Oct 19th publication: Mastering

Samuel 2k Dec 29, 2022
[NeurIPS'21] "AugMax: Adversarial Composition of Random Augmentations for Robust Training" by Haotao Wang, Chaowei Xiao, Jean Kossaifi, Zhiding Yu, Animashree Anandkumar, and Zhangyang Wang.

[NeurIPS'21] "AugMax: Adversarial Composition of Random Augmentations for Robust Training" by Haotao Wang, Chaowei Xiao, Jean Kossaifi, Zhiding Yu, Animashree Anandkumar, and Zhangyang Wang.

VITA 112 Nov 07, 2022
EigenGAN Tensorflow, EigenGAN: Layer-Wise Eigen-Learning for GANs

Gender Bangs Body Side Pose (Yaw) Lighting Smile Face Shape Lipstick Color Painting Style Pose (Yaw) Pose (Pitch) Zoom & Rotate Flush & Eye Color Mout

Zhenliang He 321 Dec 01, 2022