Skip to content

Latest commit

 

History

History
54 lines (43 loc) · 4.29 KB

README.md

File metadata and controls

54 lines (43 loc) · 4.29 KB
Raw

Semantic segmentation sample

This sample demonstrates DL model compression capabilities for semantic segmentation problem

Features:

  • UNet and ICNet with implementations as close as possible to the original papers
  • Loaders for CamVid, Cityscapes (20-class), Mapillary Vistas(20-class), Pascal VOC (reuses the loader integrated into torchvision)
  • Configuration file examples for sparsity and quantization
  • Export to ONNX compatible with OpenVINO
  • DataParallel and DistributedDataParallel modes
  • Tensorboard output

Quantize FP32 pretrained model

This scenario demonstrates quantization with fine-tuning of UNet on Mapillary Vistas dataset.

Dataset preparation

  • Obtain a copy of Mapillary Vistas train/val data here

Run semantic segmentation sample

  • If you did not install the package then add the repository root folder to the PYTHONPATH environment variable
  • Navigate to the examples/segmentation folder
  • Run the following command to start compression with fine-tuning on GPUs: python main.py -m train --config configs/unet_mapillary_int8.json --data <path_to_dataset> --weights <path_to_fp32_model_checkpoint>

It may take a few epochs to get the baseline accuracy results.

  • Use --multiprocessing-distributed flag to run in the distributed mode.
  • Use --resume flag with the path to a model from the previous experiment to resume training.
  • Use -b <number> option to specify the total batch size across GPUs

Validate your model checkpoint

To estimate the test scores of your model checkpoint use the following command: python main.py -m test --config=configs/unet_mapillary_int8.json --resume <path_to_trained_model_checkpoint> If you want to validate an FP32 model checkpoint, make sure the compression algorithm settings are empty in the configuration file or pretrained=True is set.

Export compressed model

To export trained model to ONNX format use the following command: python main.py --mode test --config configs/unet_mapillary_int8.json --data <path_to_dataset> --resume <path_to_compressed_model_checkpoint> --to-onnx unet_int8.onnx

Export to OpenVINO Intermediate Representation (IR)

To export a model to OpenVINO IR and run it using Intel Deep Learning Deployment Toolkit please refer to this tutorial.

Results

Model Compression algorithm Dataset PyTorch compressed accuracy Config path PyTorch checkpoint
UNet None CamVid 71.95 examples/semantic_segmentation/configs/unet_camvid.json Link
UNet INT8 CamVid 71.66 examples/semantic_segmentation/configs/unet_camvid_int8.json Link
UNet INT8 + Sparsity 60% (Magnitude) CamVid 71.72 examples/semantic_segmentation/configs/unet_camvid_magnitude_sparsity_int8.json Link
ICNet None CamVid 67.89 examples/semantic_segmentation/configs/icnet_camvid.json Link
ICNet INT8 CamVid 67.85 examples/semantic_segmentation/configs/icnet_camvid_int8.json Link
ICNet INT8 + Sparsity 60% (Magnitude) CamVid 67.23 examples/semantic_segmentation/configs/icnet_camvid_magnitude_sparsity_int8.json Link
UNet None Mapillary 56.23 examples/semantic_segmentation/configs/unet_mapillary.json Link
UNet INT8 Mapillary 56.1 examples/semantic_segmentation/configs/unet_mapillary_int8.json Link
UNet INT8 + Sparsity 60% (Magnitude) Mapillary 56.01 examples/semantic_segmentation/configs/unet_mapillary_magnitude_sparsity_int8.json Link