Adaptive Pyramid Context Network for Semantic Segmentation

##Easy to Run

AI Studio Link @PaddlePaddle

ackbone	Resnet101_v1c
Decode Head	APCHead
Auxiliary Head	FCNHead

---

Environment

#torch 1.8.0
conda install pytorch==1.8.0 torchvision==0.9.0 torchaudio==0.8.0 cudatoolkit=10.2 -c pytorch
pip install mmcv-full -f https://download.openmmlab.com/mmcv/dist/cu102/torch1.8.0/index.html
pip install prettytable
#paddle
python -m pip install paddlepaddle-gpu -i https://mirror.baidu.com/pypi/simple

Pretrained model：

#mmcv resnet101_v1c  mmcv/model_zoo/open_mmlab.json:
"resnet101_v1c": "https://download.openmmlab.com/pretrain/third_party/resnet101_v1c-e67eebb6.pth"

Weight tranpose：

###模型，日志,项目下载：

cd architectures
cp apcnetxxx_torch.pth architectures/pretrained/apcnetxxx_torch.pth
python torchModel2pdModel.py #generate apcnetxxx_paddle.pdparams

Ai studio

百度网盘 提取码: ttqw

Dataset：

Cityscapes:

gtFine 5000

1. Downdown to ./dataset

2. sh prepare.sh

3. cd cityscapesscripts/preparation

4. python createTrainIdLabelImgs.py

Train:

Paddle:

sh paddle.sh

Test:

Paddle:

python test.py

Measure：

Model	mIou
APCNet(paddle)	79.28%