AI Radar based video detection and development

I’m building an image- and video-driven solution whose single, clear purpose is to spot people and faces in any frame it sees. The model should automatically detect each face whether it comes from a photo, CCTV clip, webcam stream, or smartphone recording. While other object classes could be useful later, this first milestone is strictly about reliable human and facial detection. You’re free to choose the stack that suits you best—Python with OpenCV, TensorFlow, PyTorch, or a YOLO variant are all fine—as long as the final code runs on a standard GPU workstation and can be re-trained when new data arrives. I’ll provide a small starter set of annotated images; you can expand it with open-source datasets or synthetic augmentation if that helps boost robustness against lighting changes, angles, or partial occlusion. Deliverables • Trained detection model and its weights • Inference script (CLI or minimal GUI) that accepts image files and video streams and outputs frames with bounding boxes around each detected face • Brief setup & retraining guide so I can reproduce your results on my machine Acceptance is based on accurate face localisation (high precision/recall on a held-out test set) and real-time performance on 720p video. If you’ve already tackled similar projects and can turn this around efficiently, let’s talk timelines and any dataset needs right away.