OmniTrax - trained networks
YOLO Networks
- single class ant detector (trained on synthetic data)
- single class ant detector (trained on mixed synthetic and real video data from a top-down laboratory recording of leafcutter ants)
- 3 class ant detector (trained on synthetic data)
- single class termite detector (trained on synthetic data)
- 80 Class model trained on COCO
You can also use any pre-trained network from the official YOLO-model Zoo. Include the respective .obj as well as .names files and update the respective folder paths to correctly display class names.
DLC Networks
- ResNet50 ant pose-estimator (trained on mixed synthetic/real data, 10:1 ratio)
- ResNet101 ant pose-estimator (trained on mixed synthetic/real data, 10:1 ratio)
- ResNet152 ant pose-estimator (trained on mixed synthetic/real data, 10:1 ratio)
- ResNet101 (single) ant pose-estimator (trained on synthetic data)
- ResNet101 (single) stick-insect pose-estimator [full-frame] (trained on synthetic data, refined with real samples)
- ResNet101 "full_human" pose estimation converted for DeepLabCut and OmniTrax from the original DeeperCut publication
You can also make use of DeepLabCuts official Model Zoo. To use these models within OmniTrax, you will need to run the deeplabcut.export_model command. Refer to the Pose-estimation tutorial.
When using these networks trained with synthetic data, generated by replicAnt and/or our other projects in your work, please make sure to cite them:
@article{PlumLabonte2021,
title = {scAnt — An open-source platform for the creation of 3D models of arthropods (and other small objects)},
author = {Plum, Fabian and Labonte, David},
doi = {10.7717/peerj.11155},
issn = {21678359},
journal = {PeerJ},
keywords = {3D,Digitisation,Macro imaging,Morphometry,Photogrammetry,Zoology},
volume = {9},
year = {2021}
}
@misc{Plum2022,
title = {OmniTrax},
author = {Plum, Fabian},
resource = {GitHub repository},
howpublished = {https://github.com/FabianPlum/OmniTrax},
year = {2022}
}
@misc{Plumetal2023,
author = {Fabian Plum and Rene Bulla and Hendrik Beck and Natalie Imirzian and David Labonte},
title = {replicAnt: A pipeline for generating annotated images of animals in complex environments using Unreal Engine},
elocation-id = {2023.04.20.537685},
year = {2023},
doi = {10.1101/2023.04.20.537685},
publisher = {Cold Spring Harbor Laboratory},
URL = {https://www.biorxiv.org/content/10.1101/2023.04.20.537685v2},
journal = {bioRxiv}
}
© Fabian Plum, 2021 MIT License