Skip to content

README.md

Latest commit

 

History

History
16 lines (11 loc) · 580 Bytes

README.md

File metadata and controls

16 lines (11 loc) · 580 Bytes

Pytorch Model Quantization

Pose Estimation uses Pytorch for static quantization, saving, and loading of models

Get data and model

Representative Dataset: You can get it from MSCOCO val2017.zip.

Model: You can get the model from this project pytorch-pose-estimation

Quick-Start

  1. Run pth_to_int.py to get the quantized model.
  2. Run evaluate_model.py for inference.

performance

  1. Model size reduced from 200M to 50M.
  2. Inference time is reduced by about 20%.