YOLOv3 object detection vs M2Det | COCO vs Open Images v4
– YOLOv3-spp COCO (https://github.com/pjreddie/darknet)
– YOLOv3-spp Open Images v4 (https://github.com/radekosmulski/yolo_open_images)
– YOLOv3 GIoU COCO (https://github.com/generalized-iou/g-darknet)
– M2Det 512 COCO (https://github.com/qijiezhao/M2Det)
Self-explaining, I hope so.
YOLO is starting at frame 3, because of tripple buffer.
Everything was processed frame by frame, then merged into single 4K videos and after all in one 8K piece.
Thanks for the coffee if I helped you somehow:
http://bit.ly/Coffee4Karol
Already received 5 (Updated: 14th April 2019)
Thank you so much!
source