BEAGLES stands for BEhavioral Annotation and Gesture LEarning Suite, and is intended for behavioral analysis and quantification of video data. The image annotation GUI was originally forked from labelImg and the machine learning backend is based on darkflow but converts darknet configuration files to TensorFlow 2 networks.
| Branch | Status |
|---|---|
| master |   |
| dev |   |
- Darknet-style configuration files
- Automatic class balance for image classification training
- TensorFlow checkpoint files
- YOLO and VOC annotation formats
- Preconfigured to output training data to TensorBoard
- Fixed or cyclic learning rates
- Human-in-the-loop prediction and dataset expansion
- Code coverage of >60% (in progress-TF 1.x code caused setbacks)
- Automatic anchor box generation for YOLO (in progress)
- Improve maintainability to A rating (in progress)
- OBS Studio utility for USB camera arrays (in progress)
- Statistical report generation using traces (in progress)
TensorFlow 2 native codeDone!
- Tzutalin. LabelImg. Git code (2015). https://github.com/tzutalin/labelImg
- Mahmoud Aslan. Cyclic Learning Rate. Git code (2018). https://github.com/mhmoodlan/cyclic-learning-rate
- labelImg the original image annotation software BEAGLES is forked from
- darkflow the original basis of the machine learning backend
- cyclic-learning-rate the implementation of cyclic learning rates used
- traces library for non-transformative unevenly-spaced timeseries analysis
- OBS Studio video recording software
- You Only Look Once:Unified, Real-Time Object Detection
- YOLO9000: Better, Faster, Stronger
- A Framework for the Analysis of Unevenly Spaced Time Series Data
- Unevenly-spaced data is actually pretty great
- Interactive machine learning: experimental evidence for the human in the algorithmic loop
- Why Momentum Really Works
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent