Video Analytics for activity recognition

We are seeking a AI solution for video analytics focused on recognizing and analyzing human activities within video footage. The goal is to accurately identify specific activities a person is performing and calculate the time spent on each activity throughout the video. The system should be able to distinguish between predefined activities, label each occurrence accurately, and track the duration, recording start and end times to calculate total time spent on each activity. The solution should utilize Python with relevant AI and video processing libraries. A structured output, such as a summary report, should detail each detected activity with corresponding time tracking data, and ideally, a visual overlay can indicate recognized activities in real-time. Testing is required across varied video samples to ensure accuracy, with a final deliverable that includes source code, documentation, and sample output. Please share your approach, timeline, budget, and any relevant experience in video analytics or similar projects in your proposal.