🛠️ Scratchpad - CV Applications II Course by OpenCV
Computer Vision Applications using OpenCV Facial Landmark Detection Improve the speed of Facial Landmark Detection: Use face detection that comes with OS - if possible (e.g., Android, iOS) Skip frames - e.g,. run every 3 frames Downscale image, detect faces there … then propagate bboxes to original image (dividing the coordinates by the scale used for resizing the original frame.). Then just do landmark detection on original image and computed bounding box Improve landmark stabilisation Weighted Moving Average - Average the landmark point location over a small time window Kalman Filtering Optical Flow Code snippets OpenCV - Rescale an image - keep ratio import cv2 # Load image img = cv2....