Update detect_from_video.py

[mst-rajatmishra]

Hi,
I tried making some changes,

• Improved modularity by separating the logic into dedicated functions.
• Added checks for the validity of input paths.
• Enhanced comments and formatting for better understanding.
• The model loading logic is now in its own function for clarity.
• Added required=True for critical arguments to ensure necessary inputs are provided.

Thank You
RAJAT MISHRA