kaggle hubmap 2023

data download

$ kaggle datasets download -w kaerunantoka/hubmap-converted-to-coco-ds1-5fold

$ unzip hubmap-converted-to-coco-ds1-5fold.zip -d hubmap-converted-to-coco-ds1-5fold

$ rm hubmap-converted-to-coco-ds1-5fold.zip

setup mmdet 3.0.0

$ pip install torch==1.12.1+cu113 torchvision==0.13.1+cu113 --extra-index-url https://download.pytorch.org/whl/cu113

$ pip install mmcv==2.0.0rc4 -f https://download.openmmlab.com/mmcv/dist/cu113/torch1.12/index.html

$ git clone https://github.com/open-mmlab/mmdetection.git

$ pip install "mmdet==3.0.0"

$ pip install "mmcls>=1.0.0rc6"

create pseudo-labeled-coco-format-data

$ cd src
$ python create_pseudo_label_for_ds2_5fold.py -e 021 -th 0.8

$ python create_pseudo_label_for_ds3_5fold.py -e 062 -th 0.9 -dl True

$ python create_pseudo_label_for_ds2_5fold_late.py -e late001 -th 0.8

training

• default run

  $ python mmdetection/tools/train.py configs/hubmap/custom_config.py

• 5fold run (by original runner) and upload weights to kaggle dataset

  $ cd src
  $ python train_5fold.py configs/hubmap/exp038.py

• Multi GPU run

  
  NNODES=${NNODES:-1}
  NODE_RANK=${NODE_RANK:-0}
  PORT=${PORT:-29500}
  MASTER_ADDR=${MASTER_ADDR:-"127.0.0.1"}

PYTHONPATH="$(dirname $0)/..":$PYTHONPATH \ python -m torch.distributed.launch \ --nnodes=$NNODES \ --node_rank=$NODE_RANK \ --master_addr=$MASTER_ADDR \ --nproc_per_node=2 \ --master_port=$PORT \ train_5fold_ddp_late.py \ configs/hubmap/late007.py \ --launcher pytorch ${@:3}

## show mean mAP

single case

$ python show_mean_map_from_mmdet_log.py -e exp074 -p work_dirs $ python show_mean_map_from_mmdet_log.py -e exp078 -p /external_disk/work_dirs $ python show_mean_map_from_mmdet_log.py -e late001 -p work_dirs

multiple case

$ python show_mean_map_from_mmdet_log.py -e 'exp021 exp027' -p work_dirs

## upload output

$ kaggle datasets create --dir-mode zip -p /external_disk/work_dirs/exp021

$ kaggle datasets version --dir-mode zip -p /external_disk/work_dirs/exp065 -m "reduce pth"

## linter and formatter

$ make source-code-format