This work was presented at the "Computer Vision with Small Data: A Focus on Infants and Endangered Animals CV4Smalls" workshop at WACV2024 and won the best paper award.
Code has been used with CUDA 11.0, CudNN 8.3.2, Python 3.8 on Ubuntu 18.04 Please execute the following:
git clone https://github.com/mshooter/DigiDogs_release.git
cd DigiDogs_release
conda env create -f environment.yaml
conda activate digidogs
pip install -r requirements.txt
# this will install the src code called digidogs
pip install . Download the data from the following link and modify the file to the folder in the config files in digidogs/configs/*.yaml or follow the folder structure.
The folder structure should look like this, so if directories are not created after git cloning, then please do create them e.g. logs, checkpoints:
DigiDogs_release
├── checkpoints
│ └── ...
├── logs
│ └── ...
├── configs
│ ├── train_digidogs.yaml
│ ├── test_digidogs.yaml
│ ├── test_stanext.yaml
├── data
│ ├── \*.png
├── scripts
│ └── demo.py
│ └── launch.py
├── digidogs
│ └── configs
│ └── datasets
│ └── models
│ └── rgbddog
│ └── utilsWe are showing our method on internet images. We already have put images in the folder, but if you want to do inference on your images: 1. Please put your images in the folder called data/ The argument is: the image path to images to the images you want to apply the method to Argument "custom" needs to be put, indicating it is custom images.. For example:
python script/demo.py custom DigiDogs_release/data/If you need to change the checkpoints, there are several checkpoints available in the checkpoint folder:
- one trained with the DigiDogs dataset (gta)
- one trained with the RGBDDog dataset (rgbd)
- one trained with the SVM dataset (svm) Please change line #64 in scripts/demo.py
The RGBD-Dog and the GTA will return a complete skeleton, while the svm will only return some keypoints from one point of view.
- Check inside the bash files how we run the script, for training and testing
- Please change the "data_dir" variable in the *.yaml files, to where your dataset is located
- If you want to adapt parameters etc. you can modify the config file in digidogs/configs/ and the parameters in the bash files
Bash train.sh
If you are having the below error:
AttributeError: partially initialized module 'cv2' has no attribute 'gapi_wip_gst_GStreamerPipeline' (most likely due to a circular import)
Please do the following:
pip uninstall opencv-python
pip uninstall opencv-contrib-python
pip uninstall opencv-python-headless
pip install opencv-python
pip install opencv-contrib-python
pip install opencv-python-headless
- Release test code.
@InProceedings{Shooter_2024_WACV,
author = {Shooter, Moira and Malleson, Charles and Hilton, Adrian},
title = {DigiDogs: Single-View 3D Pose Estimation of Dogs Using Synthetic Training Data},
booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) Workshops},
month = {January},
year = {2024},
pages = {80-89}
}
Please refer to the license file or below.
Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC BY-NC-SA 4.0)
Copyright (c) 2025 Moira Shooter, Charles Malleson, Adrian Hilton, University of Surrey
This work is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. To view a copy of this license, visit http://creativecommons.org/licenses/by-nc-sa/4.0/
BY: credit must be given to the creator(s). NC: Only noncommercial uses of the work are permitted. SA: Adaptations must be shared under the same terms.
Note If you have any questions regarding the licensing please contact either Professor Adrian Hilton or Dr Charles Malleson.
Please contact Moira Shooter for enquiries