FastDeploy/examples/vision/segmentation/paddleseg

English | 简体中文

PaddleSeg Model Deployment

Model Description

• PaddleSeg develop

FastDeploy currently supports the deployment of the following models

• U-Net models
• PP-LiteSeg models
• PP-HumanSeg models
• FCN models
• DeepLabV3 models

【Attention】For PP-Matting、PP-HumanMatting and ModNet deployment, please refer to Matting Model Deployment

Prepare PaddleSeg Deployment Model

For the export of the PaddleSeg model, refer to Model Export for more information

Attention

• The exported PaddleSeg model contains three files, including model.pdmodel、model.pdiparams and deploy.yaml. FastDeploy will get the pre-processing information for inference from yaml files.

Download Pre-trained Model

For developers' testing, part of the PaddleSeg exported models are provided below.

• without-argmax export mode: Not specified--input_shape，specified--output_op none
• with-argmax export mode：Not specified--input_shape，specified--output_op argmax

Developers can download directly.

Model	Parameter Size	Input Shape	mIoU	mIoU (flip)	mIoU (ms+flip)
Unet-cityscapes-with-argmax | Unet-cityscapes-without-argmax	52MB	1024x512	65.00%	66.02%	66.89%
PP-LiteSeg-B(STDC2)-cityscapes-with-argmax | PP-LiteSeg-B(STDC2)-cityscapes-without-argmax	31MB	1024x512	79.04%	79.52%	79.85%
PP-HumanSegV1-Lite-with-argmax(General Portrait Segmentation Model) | PP-HumanSegV1-Lite-without-argmax(General Portrait Segmentation Model)	543KB	192x192	86.2%	-	-
PP-HumanSegV2-Lite-with-argmax(General Portrait Segmentation Model) | PP-HumanSegV2-Lite-without-argmax(General Portrait Segmentation Model)	12MB	192x192	92.52%	-	-
PP-HumanSegV2-Mobile-with-argmax(General Portrait Segmentation Model) | PP-HumanSegV2-Mobile-without-argmax(General Portrait Segmentation Model)	29MB	192x192	93.13%	-	-
PP-HumanSegV1-Server-with-argmax(General Portrait Segmentation Model) | PP-HumanSegV1-Server-without-argmax(General Portrait Segmentation Model)	103MB	512x512	96.47%	-	-
Portait-PP-HumanSegV2-Lite-with-argmax(Portrait Segmentation Model) | Portait-PP-HumanSegV2-Lite-without-argmax(Portrait Segmentation Model)	3.6M	256x144	96.63%	-	-
FCN-HRNet-W18-cityscapes-with-argmax | FCN-HRNet-W18-cityscapes-without-argmax(GPU inference for ONNXRuntime is not supported now)	37MB	1024x512	78.97%	79.49%	79.74%
Deeplabv3-ResNet101-OS8-cityscapes-with-argmax | Deeplabv3-ResNet101-OS8-cityscapes-without-argmax	150MB	1024x512	79.90%	80.22%	80.47%

Detailed Deployment Tutorials

• Python Deployment
• C++ Deployment