This is a test of YOLOv2 on KITTI. YOLOv2 is the state of the art object detector. And it is also amazingly fast and easy to use.
* I am not the author of YOLO. I just use it to train on KITTI.

YOLOv2: [ Ссылка ]
KITTI: [ Ссылка ]

Input Image Size = 1238 x 374 (KITTI data_tracking_image_2/training/image_02/0019)
Numbers of Frames = 1057
Network Input Size = 1280 x 384 (set close to input image size)
Measured Speed = 24.1 fps (including all processing time)
Hardware = i7-7700K / 32G DDR4 / GTX1070 / SSD
OS = Ubuntu 16.04 64bit

свернуть