PPLiteSeg.pytorch
pytorch of the SOTA real-time segmentation network ppliteseg
| Model | Backbone | Training Iters | Train Resolution | Test Resolution | mIoU | mIoU (flip) | mIoU (ms+flip) |
|---|---|---|---|---|---|---|---|
| PP-LiteSeg-T | STDC1 | 160000 | 1024x512 | 1025x512 | 73.10% | 73.89% | - |
| PP-LiteSeg-T | STDC1 | 160000 | 1024x512 | 1536x768 | 76.03% | 76.74% | - |
| PP-LiteSeg-T | STDC1 | 160000 | 1024x512 | 2048x1024 | 77.04% | 77.73% | 77.46% |
| PP-LiteSeg-B | STDC2 | 160000 | 1024x512 | 1024x512 | 75.25% | 75.65% | - |
| PP-LiteSeg-B | STDC2 | 160000 | 1024x512 | 1536x768 | 78.75% | 79.23% | - |
| PP-LiteSeg-B | STDC2 | 160000 | 1024x512 | 2048x1024 | 79.04% | 79.52% | 79.85% |
here we convert the model and weights of PP-LiteSeg-B(1024x512) from paddlepaddle to torch.
Model&Weight
pp_liteseg.py : pytorch model
ppliteset_pp2torch_cityscape_pretrained.pth: the cityscape pretrained weights trained with paddleseg
demo of paddleseg:
demo of pytorch:
Difference
upsample
PaddleSeg use "bilinear" mode, while in pytorch, I use "nereast" mode in order to convert to TensorRT .
BatchNorm
paddleseg :momentum=0.9
while the default setting of torch: momentum=0.1.
train
use ddrnet to train this model, set the coefficient of three seghead(1/8,1/16and1/32) loss to 1 while training.
Demo
see demo.py