This project provides aboundant choices of quantization strategies (such as the quantization algoirthms, training schedules and empirical tricks) for quantizing the image classification neural networks into low bit counterparts. Associated projects demonstrate that this project can also benefit other computer vision tasks, such as object detection, segmentation and text parsing. Pretrained models are provided to show high standard of the code on achiving appealing quantization performance.
- Classification: Performance
- Detection: Performance
- Segmentation: Performance
- Text parsing: Performance
- Text parsing (preparing)
- 2020.07.08 Instance Segmentation
- 2020.07.08 Object Detection
- 2020.06.23 Add classification quantization
Please cite the following work if you find the project helpful.
Also Cite correspoding publication when you choose dedicated algorithms.
Current Contributors:
To contribute, PR is appreciated and suggestions are welcome to discuss. Private conact is available at blueardour@gmail.com
For academic use, this project is licensed under the 2-clause BSD License. See LICENSE file.