Hi! My name is Konstantin. I'm experienced developer and researcher in the Computer Vision, Object Detection and Machine Learning areas. I have done many of projects more or less similar to your, including, for example, football players and ball position capturing, intruders tracking for CCTV, aerial photo stitching, diseases localization in medical images, etc. All these projects were built with real-time and detection performance in mind with widely usage of OpenCL and CUDA libraries in time-critical computation parts. So, I think I'm right person for this task. But as in all similar projects, the complexity of exact CV task highly depends of particular requirements, type of objects to be detected, limitations, etc. Can you please provide more details about the task, objects, environment, etc.?
With best regards, Konstantin