Computer vision capstones are some of the most visually demoable projects you can build — a live camera feed with bounding boxes wins over a panel fast. These ideas are grouped from classic image classification to real-time detection.
YOLO/MobileNet-based detector on a webcam feed — a proven, easy-to-explain capstone from the pandemic era that still demos well.
Face-encoding-based recognition marking attendance automatically from a classroom camera.
Detection plus OCR pipeline for reading plates from images/video, useful for parking or toll use cases.
OCR pipeline converting scanned handwritten notes into searchable text.
YOLO-based detection counting/identifying products on a shelf for stock-out alerts.
Eye-aspect-ratio or CNN-based drowsiness detection from a dashboard-facing camera, with an audio alert.
Hand-gesture classification translating sign language into text/speech in real time.
A CNN classifier identifying disease from leaf photos — pairs well with a mobile app front-end for field use.
CNN+RNN/transformer pipeline generating natural-language captions for input images.