HDNet: Human Depth Estimation For Multi-Person Camera-Space Localization

This repository includes the source code for our ECCV 2020 paper on monocular image human depth estimation for camera-space multi-person 3D pose localization. Please read our paper for more details at https://arxiv.org/abs/2007.08943.

Bibtex:

@inproceedings{lin2020hdnet,
    title={HDNet: Human Depth Estimation for Multi-Person Camera-Space Localization},
    author={Lin, Jiahao and Lee, Gim Hee},
    booktitle={ECCV},
    year={2020}
}

Environment

The code is developed and tested on the following environment

• Python 3.5.2
• PyTorch 1.1.0
• CUDA 9.0

Dataset

The source code is for training and testing on Human3.6M dataset.

Training data

To generate images for training, we provide a script generate_training_images.py to transform raw videos to full resolution images with proper naming. Please download raw videos from Human3.6M website. After running the provided script, put the generated train folder under data/Human36M/images/ directory.

Testing data

Testing images can be downloaded here. Put the test folder under data/Human36M/images/ directory.

Annotations

Processed annotations can be downloaded here. Put all files under data/Human36M/annotations/ directory.

The dataset should be organized as shown below:

    data/Human36M
            └── images
                └── train
                    ├── s_01_act_02_subact_01_ca_01/
                    ├── ...
                    └── s_08_act_16_subact_02_ca_04/
                └── test
                    ├── s_09_act_02_subact_01_ca_01/
                    ├── ...
                    └── s_11_act_16_subact_02_ca_04/
            └── annotations
                ├── subject_1.h5
                ├── subject_1_cpn.h5
                ├── ...
                ├── subject_11.h5
                └── subject_11_cpn.h5

Training

To train a model, run:

python train.py [OPTIONS value]

Refer to the source code in train.py for available OPTIONS.

Logs and models will be saved under exp/[EXP_TAG].

Evaluation

To evaluate a model named EXP_TAG, run:

python test.py --tag EXP_TAG

Our pre-trained model is available for download here. Place the model under exp/pretrained/saved_models/ and run:

python test.py --tag pretrained