MMSegmentation is an open source object segmentation toolbox based on PyTorch. It is a part of the OpenMMLab project.
Please refer to get_started.md for installation.
model | OnnxRuntime | TensorRT | NCNN | PPLNN | OpenVino | model config file(example) |
---|---|---|---|---|---|---|
FCN | Y | Y | Y | Y | Y | ${MMSEG_DIR}/configs/fcn/fcn_r50-d8_512x1024_40k_cityscapes.py |
PSPNet* | Y | Y | Y | Y | Y | ${MMSEG_DIR}/configs/pspnet/pspnet_r50-d8_512x1024_40k_cityscapes.py |
DeepLabV3 | Y | Y | Y | Y | Y | ${MMSEG_DIR}/configs/deeplabv3/deeplabv3_r50-d8_512x1024_40k_cityscapes.py |
DeepLabV3+ | Y | Y | Y | Y | Y | ${MMSEG_DIR}/configs/deeplabv3plus/deeplabv3plus_r50-d8_512x1024_40k_cityscapes.py |
Fast-SCNN* | Y | Y | N | Y | Y | ${MMSEG_DIR}/configs/fastscnn/fast_scnn_lr0.12_8x4_160k_cityscapes.py |
-
Only
whole
inference mode is supported for all mmseg models. -
PSPNet, Fast-SCNN only support static shape, because nn.AdaptiveAvgPool2d is not supported in most of backends dynamically.
-
For models only supporting static shape, you should use the deployment config file of static shape such as
configs/mmseg/segmentation_tensorrt_static-1024x2048.py
.
None