Train¶
# train
CUDA_VISIBLE_DEVICES=0 PORT=29500 \
./tools/dist_train.sh configs/rretinanet/rretinanet_obb_r50_fpn_1x_dota_v3.py 1
Test¶
CUDA_VISIBLE_DEVICES=0 PORT=29500 \
./tools/dist_test.sh configs/rretinanet/rretinanet_obb_r50_fpn_1x_dota_v3.py \
work_dirs/rretinanet_obb_r50_fpn_1x_dota_v3/epoch_12.pth 1 --mAP
Inference & Submit¶
CUDA_VISIBLE_DEVICES=0 PORT=29500 \
./tools/dist_test.sh configs/rretinanet/rretinanet_obb_r50_fpn_1x_dota_v3.py \
work_dirs/rretinanet_obb_r50_fpn_1x_dota_v3/epoch_12.pth 1 --format-only\
--eval-options submission_dir=work_dirs/rretinanet_obb_r50_fpn_1x_dota_v3/Task1_results
Crop Images¶
For DOTA dataset, please crop the original images into 1024×1024 patches with an overlap of 200 by run
python tools/split/img_split.py --base_json \
tools/split/split_configs/split_configs/dota1_0/ss_trainval.json
python tools/split/img_split.py --base_json \
tools/split/split_configs/dota1_0/ss_test.json
Please change path in ss_trainval.json
, ss_test.json
to your path. (Forked from BboxToolkit, which is faster then DOTA_Devkit.)