The solution is supposed to take in videos and analyze each video at a given frame rate say 5 frames per second. This is what should happen in analysis - for each frame of a video (at the given frame rate), a state-of-the-art deep learning based object detector (for example YOLOv3) should be used to detect several objects like man, woman, face, child, vehicle, bus, truck, car, bicycle, motorbike etc. and for each detected object, it should extract features/embeddings using, say AlexNet and storing them to an efficient persistent store with appropriate tagging to be able to retrieve the analysis results efficiently for any given frame of any given video. The important thing is that the hardware of choice for this "solution" should be such that 24 hours of videos can be analyzed (as above) in 10 mins. Also, the solution should be developed as a module and it should be possible to integrate this module with a larger system through well defined APIs.
14 фрилансеров(-а) в среднем готовы выполнить эту работу за $519
Hi. I have review your project description. I have full experiences in Development. I am very interested in your project and also ready to start work immediately. Hoping for your soon Reply. Thank You Regard.
Hi, I am professional in Image processing and Machine learning. I analyzed the video deep learning based object detector previously. I can do your project perfectly. Thanks.