Integrate Yolo to initiate the SAMURAI

[sendtohen]

I would love to use Yolo to initiate the SAMURAI. I tried, but got errors.
SAMURAI mode: False
or
I would get
File "C:\Users\Documents\samurai\sam2\sam2\modeling\sam2_base.py", line 665, in _prepare_memory_conditioned_features
iou_score = output_dict["non_cond_frame_outputs"][i]["best_iou_score"] # Get mask affinity score
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^
KeyError: 17

[hsiangwei0903]

can you provide more details on your modifications?

[sendtohen]

Yolo_sam.txt
Error
error.txt
Sorry still fairly new to coding as a whole, as I am fascinated by what you are doing.

[hsiangwei0903]

I think your code will modify the frame_idx to 17, which can further caused error in the following line. I would suggest write a separate script to detect the ball and save it into txt, and then run our code for tracking.

_, _, masks = predictor.add_new_points_or_box(state, box=bbox, frame_idx=frame_idx, obj_id=0)

[sendtohen]

thanks for the suggestion, i will give this method a try.