YOLO: real-time object detection

Course Project under course CS698N - Recent Advances in Computer Vision

• Followed CVPR’16 paper ”You only look once: Unified, real-time object detection” by Joseph Redmon et al.(http://pjreddie.com/media/files/papers/yolo_1.pdf), training and testing YOLO and tiny-YOLO CNN models on PASCAL VOC 2012 Object Detection Dataset
• Trained and obtained detection results on the object detection benchmark KITTI by using a YOLO network pre-trained on ImageNet
• Did an extensive parameter study on YOLO, studying the changes observed on increasing/decreasing the grid resolution and increasing number of bounding boxes per grid cell
• Our trained Tiny-YOLO model runs at around 50-70 fps on NVIDIA GeForce GTX 760 GPU with mAP 44.65% on PASCAL VOC 2012 validation dataset

Check out the detailed report here and presentation here.

YOLO : http://pjreddie.com/darknet/yolo/