search

Tramac / mobilenetv3-segmentation

MobileNetV3 for Semantic Segmentation.
Apache License 2.0
97 stars 20 forks source link

Image size #4

Open bluesky314 opened 4 years ago

bluesky314 commented 4 years ago

Why is it requiring such huge images of >512? Original mobilenetV3 accepts small 244*244 images.

Tramac commented 4 years ago

Please change it if necessary.

dscarmo commented 3 years ago

The segmentation head expects big images (from datasets like Cityscapes). Check the LRASPP implementation in the paper. I was able to fix this by using AdaptiveAvgPooling instead of AvgPool in the LRASPP implementation.

sungh66 commented 1 year ago

The segmentation head expects big images (from datasets like Cityscapes). Check the LRASPP implementation in the paper. I was able to fix this by using AdaptiveAvgPooling instead of AvgPool in the LRASPP implementation.

When i exported ONNX model, ONNX cannot support adaptive_avg_pool2d.How can i fix the problem?Or when i export PNNX model ,PNNX cannot support node Tensor.expand_as.