Help request for testing setup

[alesiavq]

Hello Adrien Meyer and team,

Thank you for this great tool. I am currently trying to run Ultrasam on a custom dataset to segment the fetal femur bone from a US image. we followed your instructions and used the following command line to run. However, there are no segmentations generated. Can you kindly guide us on how we can run Ultrasam to generate bone segmentation? This is the command we used.

mim test mmdet configs/UltraSAM/UltraSAM_full/UltraSAM_box_refine.py --checkpoint UltraSam.pth --cfg-options test_dataloader.dataset.data_root="test_data" test_dataloader.dataset.ann_file="test_data/sample_COCO.json" test_dataloader.dataset.data_prefix.img="test_data/sample_images" test_evaluator.ann_file="test_data/test_data/sample_COCO.json" --work-dir ./work_dir/example --show-dir ./show_dir

I have shared our sample_coco.json file and sample image here. Which file should we be running, and how can we generate the bbox in this case? If you could please point us in the right direction, it would be greatly appreciated.

Thanks!

sample_COCO.json

[Chop1]

Hi, looking at your .json, my guess is that the segmentation field is not encoded in RLE.
mim test expect the GT in the annotation file, as it also compute the metrics.

[alesiavq]

Hi, looking at your .json, my guess is that the segmentation field is not encoded in RLE. mim test expect the GT in the annotation file, as it also compute the metrics.

Hello,

Thank you for your response. If we do not have the ground truth and want to output the segmentation, what would you recommend we try. Thank you!