New ideas, thought about needed features will be stored in this file.
- Darknet YOLO v3 layers
- Convolution layer;
- Maxpool layer;
- Upsample layer;
- Route layer;
- Shortcut layer;
- YOLO layer;
- Utilities
- Darknet-based weights parser
- Darknet-based neural network configurtion file parser
- NMS for detected objects and all corresponding functions such as IOU, bbox rectifying
- Convert image to slice of float32
- Resize image and average color picker
- Pixel extracting and adapting it for []float32
- New Gorgonia operations (or modified existing ones):
- Upsample and its derivative
- YOLO and its derivative
- Test coverage (Its always good)
- Full inference example
- Training WIP
- Loss function WIP, PRs are welcome
- Proper backpropagation WIP, PRs are welcome
- Optimizations (replace 'raw loop' in code with Gorgonia core functions)
- GPU (Here we have much work to do)
- Instructions
- Code itself (with Gorgonia CUDA-based library)
- Benchmarks
- Documentation (fill README.md with usefull stuff) WIP
Updated at: 2020-10-28