ContextQMD
Back to Unilm

ADE20k Semantic segmentation with BEiT

beit/semantic_segmentation/README.md

latest2.7 KB
Original Source

ADE20k Semantic segmentation with BEiT

Getting started

  1. Install the mmsegmentation library and some required packages.
bash
pip install mmcv-full==1.3.0 mmsegmentation==0.11.0
pip install scipy timm==0.3.2
  1. Install apex for mixed-precision training
bash
git clone https://github.com/NVIDIA/apex
cd apex
pip install -v --disable-pip-version-check --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" ./
  1. Follow the guide in mmseg to prepare the ADE20k dataset.

Fine-tuning

Command format:

tools/dist_train.sh <CONFIG_PATH> <NUM_GPUS>  --work-dir <SAVE_PATH> --seed 0  --deterministic --options model.pretrained=<IMAGENET_CHECKPOINT_PATH/URL>

For example, using a BEiT-base backbone with UperNet:

bash
bash tools/dist_train.sh \
    configs/beit/upernet/upernet_beit_base_12_640_slide_160k_ade20k_pt2ft.py 8 \
    --work-dir /path/to/save --seed 0  --deterministic \
    --options model.pretrained=https://github.com/addf400/files/releases/download/v1.0/beit_base_patch16_224_pt22k_ft22k.pth

More config files can be found at configs/beit/upernet.

Evaluation

Command format:

tools/dist_test.sh  <CONFIG_PATH> <CHECKPOINT_PATH> <NUM_GPUS> --eval mIoU

For example, evaluate a BEiT-base backbone with UperNet:

bash
bash tools/dist_test.sh configs/beit/upernet/upernet_beit_base_12_640_slide_160k_ade20k_pt2ft.py \ 
    https://github.com/addf400/files/releases/download/v1.0/beit_base_patch16_640_pt22k_ft22ktoade20k.pth  4 --eval mIoU

Expected results:

+--------+-------+-------+-------+
| Scope  | mIoU  | mAcc  | aAcc  |
+--------+-------+-------+-------+
| global | 53.61 | 64.82 | 84.62 |
+--------+-------+-------+-------+

Multi-scale + flip (\*_ms.py)

bash tools/dist_test.sh configs/beit/upernet/upernet_beit_base_12_640_slide_160k_ade20k_ms.py \
    https://github.com/addf400/files/releases/download/v1.0/beit_base_patch16_640_pt22k_ft22ktoade20k.pth  4 --eval mIoU

Expected results:

+--------+-------+-------+------+
| Scope  | mIoU  | mAcc  | aAcc |
+--------+-------+-------+------+
| global | 54.26 | 65.28 | 84.9 |
+--------+-------+-------+------+

Acknowledgment

This code is built using the mmsegmentation library, Timm library, the Swin repository, XCiT and the SETR repository.