https://github.com/jahongir7174/sixdrepnet
6D Rotation Representation for Unconstrained Head Pose Estimation
https://github.com/jahongir7174/sixdrepnet
head-pose-estimation opencv-python python pytorch
Last synced: 3 months ago
JSON representation
6D Rotation Representation for Unconstrained Head Pose Estimation
- Host: GitHub
- URL: https://github.com/jahongir7174/sixdrepnet
- Owner: jahongir7174
- Created: 2024-02-28T00:12:30.000Z (over 1 year ago)
- Default Branch: master
- Last Pushed: 2024-03-27T22:26:04.000Z (over 1 year ago)
- Last Synced: 2025-03-24T12:46:59.149Z (7 months ago)
- Topics: head-pose-estimation, opencv-python, python, pytorch
- Language: Python
- Homepage: https://arxiv.org/pdf/2202.12555.pdf
- Size: 82.2 MB
- Stars: 7
- Watchers: 0
- Forks: 1
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
6D Rotation Representation for Unconstrained Head Pose Estimation
### Note
* The default train dataset is `300W-LP`
* The default test dataset is `AFLW2000`
### Installation
```
conda create -n PyTorch python=3.11
conda activate PyTorch
conda install pytorch torchvision torchaudio pytorch-cuda=11.8 -c pytorch -c nvidia
pip install opencv-python==4.5.5.64
pip install scipy
pip install tqdm
pip install timm
```
### Train
* Configure your dataset path in `main.py` for training
* Download [IMAGENET](https://github.com/jahongir7174/SixDRepNet/releases/tag/v0.0.1) pretrained weights
* Run `python main.py --train` for Single-GPU training
* Run `bash main.sh $ --train` for Multi-GPU training, `$` is number of GPUs
### Test
* Configure your dataset path in `main.py` for testing
* Run `python main.py --test` for testing
### Demo
* Configure your video path in `main.py` for visualizing the demo
* Run `python main.py --demo` for demo
### Results
| Backbone | Epochs | Pitch | Yaw | Roll | MAE | Parameters(M) | FLOPS (B) | Throughput (images/s) |
|:------------:|:------:|------:|-----:|-----:|-----:|--------------:|----------:|----------------------:|
| RepNet-A2 | 90 | 4.78 | 3.68 | 3.25 | 3.90 | 25.49 | 5.1 | 1322 |
| RepNet-B1G2* | 30 | 4.91 | 3.63 | 3.37 | 3.97 | 41.36 | 8.8 | 792 |
`*` means that the results are from original repo, see reference
#### Reference
* https://github.com/thohemp/6DRepNet