Regarding the code for combining the broken boxes

[shreejalt]

Hi,
Thanks for your contribution. The approach seems to be perfect. I wanted to know if you can help me in finding the code for correcting and combining the boxes

Thanks,
Shreejal

[previtus]

Hi Shreejal,
Thank you for your interest. I believe this was done with functions in: https://github.com/previtus/AttentionPipeline/blob/master/video_parser_v2/processing_code/bbox_postprocessing.py

[shreejalt]

Thanks a lot, @previtus

I wanted to ask small questions regarding your project.

I wanted to use different object detectors with different input resolutions. So I was seeing the code and was curious to know a few things.

1. How can I incorporate the box post-processing for vehicle class? It seems you have taken into account just the person class. Also, how can I change the code so that I can take into consideration square boxes(Not vertical as mentioned in your paper) to merge the broken boxes?
2. What are the main functions which I can use if I don't want to use threading and parallel processing (Also no client-server architecture)
3. I have seen that there is a main_loop.py file. Can I use only those functions and the functions called from CropCoordinates and AttentionModule to complete the code with different resolutions?

I really appreciate your prompt response.
Thanking you in advance

Shreejal

[shreejalt]

Hi @previtus
Let me know your thoughts on the same