kozzy / model-quantization

Collections of model quantization algorithms

QTool: A low bit quantization toolbox for computer vision neural networks

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.

Supported Task

• Classification: Performance
• Detection: Performance
• Segmentation: Performance
• Text parsing: Performance

Update History

• Text parsing (preparing)
• 2020.07.08 Instance Segmentation
• 2020.07.08 Object Detection
• 2020.06.23 Add classification quantization

Citation

Please cite the following work if you find the project helpful.

Also Cite correspoding publication when you choose dedicated algorithms.

Contributors

Current Contributors:

• Blueardour
• bohanzhuang
• liujingcs
• liuchunlei0430

To contribute, PR is appreciated and suggestions are welcome to discuss. Private conact is available at [email protected]

License

For academic use, this project is licensed under the 2-clause BSD License. See LICENSE file.