deeplabv1-resnet38's Introduction

DeeplabV1 ResNet38 Retrain

First download the ResNet38 pretrain from here.

Training

Please modify the configration in config.py according to your device firstly.

CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.launch --nproc_per_node 2 tools/train.py

Test

Don't forget to check test configration in config.py first.

VOC12

CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.launch --nproc_per_node 2 tools/test.py --period val --test_ckpt [checkpoint path] --test_flip --test_multiscale --test_save prob_npy

COCO14

CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.launch --nproc_per_node 2 tools/test.py --period val --test_ckpt [checkpoint path] --test_flip --test_multiscale --test_save prob_npy

CRF Postprocess

VOC12

python tools/make_crf.py \
--list val.txt \
--data-path data \
--predict-dir prob_npy \
--predict-png-dir pred_png \
--num-cls 21 \
--dataset voc12

COCO14

python tools/make_crf.py \
--list val.txt \
--data-path data \
--predict-dir prob_npy \
--predict-png-dir pred_png \
--num-cls 91 \
--dataset coco

deeplabv1-resnet38's People

Contributors

Stargazers

Watchers

deeplabv1-resnet38's Issues

About MCTformer

Can you provide instructions for all the training steps? I followed the author's instructions and the final segmentation result of mIoU was only 67.5%.

Recommend Projects