Semantic Segmentation

Segmentation on ADE20K is implemented based on MMSegmentation.

Models

Model	mIoU	Log
LSNet-T	40.1	lsnet_t_semfpn.json
LSNet-S	41.6	lsnet_s_semfpn.json
LSNet-B	43.1	lsnet_b_semfpn.json

Requirements

pip install mmsegmentation==0.30.0

Data preparation

Please prepare ADE20K dataset following insructions in MMSeg. The data should appear as:

├── segmentation
│   ├── data
│   │   ├── ade
│   │   │   ├── ADEChallengeData2016
│   │   │   │   ├── annotations
│   │   │   │   │   ├── training
│   │   │   │   │   ├── validation
│   │   │   │   ├── images
│   │   │   │   │   ├── training
│   │   │   │   │   ├── validation

Testing

./tools/dist_test.sh configs/sem_fpn/fpn_lsnet_b_ade20k_40k.py pretrain/lsnet_b_semfpn.pth 8 --eval mIoU

Training

Download ImageNet-1K pretrained weights into ./pretrain

./tools/dist_train.sh configs/sem_fpn/fpn_lsnet_b_ade20k_40k.py 8 --seed 0 --deterministic