lwl algorithm seems not to support run_video.py?

[Ying-Kang]

I just tried to run "run_video.py" with lwl using lwl_boxinit but raise error: KeyError: previous_output

After debuging, I found that info is None (there is no doubt tha keyerror appear with empty dict) after ROI is selected, which means ROI initialization is not success for lwl algo in run_video.py

when i switch pattern into webcam with lwl, this kind of error disappeared.

It is obvious that "run_webcam.py" file provided thourough usage of info variable while "run_video.py" not

could you please fulfill the run_video file with lwl?

i'd appreciate it if you could help

thx anyway

[goutamgmb]

Hi,

Yes, that is true. Seems we haven't updated run_video script to support lwl. Thanks for notifying the problem. We will soon update the script.

[Ying-Kang]

I made my version by imitating the coding style in "run_webcam.py" Couldn't wait to see your version int the future

[Ahsanr312]

Can you share your run_video.py file ? As it's still not working for LWL

[Mprjct]

but it doesn't have "box_label_encoder" yet

[mattpfr]

@Ying-Kang @Ahsanr312 @Mprjct run_video.py should support LWL but you have to run with lwl_boxinit, as LWL requires a segmentation mask to start, which is not supported. Alternatively, you can use RTS that will also generate a segmentation mask based on the bounding boxes you draw.

Could you please let me know if it works for you so we can close that ticket ?