Closed iwander-all closed 3 months ago
After efforts, I have worked out the test demo. If you want to run
./tools/dist_test.sh ./projects/configs/vma/vma_icurb.py ./path/to/ckpts.pth 8
,
the command should be:
./tools/dist_test.sh ./projects/configs/vma_icurb.py ./ckpts/icurb.pth 8
change dist_test.sh with this:
PYTHONPATH="$(dirname $0)/..":$PYTHONPATH \ python -m torch.distributed.run --nproc_per_node=$GPUS --master_port=$PORT \ $(dirname "$0")/test.py $CONFIG $CHECKPOINT --launcher pytorch ${@:4} --eval icurb --eval-option show=True show_dir=$(dirname $0)/../visualize_result/
,
and in icurb_dataset.py , delete first line in func def show_result(pred_points, out_dir, image_path)
like this:
# image_path = image_path.replace('tiff', 'jpg')
for this demo, the command is:
python ./demo/custom_infer.py ./demo/image_data/ ./projects/configs/vma_res152_e80_line.py ./ckpts/sd_line.pth --trajectory_sample_num 15 --element_type line --out_dir ./demo/out/ --visualize True
python ./demo/custom_infer.py ./demo/image_data/ ./projects/configs/vma_res152_e80_box.py ./ckpts/sd_box.pth --trajectory_sample_num 15 --element_type box --out_dir ./demo/out/ --visualize True
python ./demo/custom_infer.py ./demo/image_data/ ./projects/configs/vma_icurb_backbone_e80_freespace.py ./ckpts/sd_freespace.pth --trajectory_sample_num 15 --element_type freespace --out_dir ./demo/out/ --visualize True
I donot know why trajectory is needed in this demo.
you offer commands like :
but what is {root_directory} {config_path} {checkpoint_path} ?