PRAVEEN KUMAR

1. Cash box monitoring system for the large store. With help of CCTV footage, selected the ROI of the cashbox and selected one closed box and open box image and applied SSIM on that ROI, so whenever the box was opened recording starts and box was closed recording stops. 2. ANPR (Automatic Number plate Recognition): Built models to detect vehicles violating traffic rules, detect the number plate, and read the number plate for the Automatic Challan system using yolo, cnn classification, ocr, and open cv techniques, achieving almost 90% accuracy. 3. Surveillance system for manufacturing industries and warehouse: Built models for PPE, Mask, Helmet, and Shoes detection for safety alerts (yolo model, accuracy > 98%). Also implemented Intrusion detection, container counting using chain pulley movement, damaged box detection, and social distancing detection. 4. Demographics at Store: Classifying the person present in store based on Gender, Age group, and Employee status. 5. HAR (Human Activity Recognition): Classifying Human Activity into different classes based on video data, built using person detection, keypoint extraction, and training an LSTM model on keypoints for each person. 6. Human Help at Airport: Developed a system to check if a person is carrying more than 3 pieces of luggage, and implemented alerts for old persons or pregnant ladies at security check-in, along with human fall prediction.