YOLOv3 object detection vs M2Det  |  COCO vs Open Images v4 @KarolMajek
YOLOv3 object detection vs M2Det  |  COCO vs Open Images v4  @KarolMajek
Uploaded April 2019 | Updated September 2026, 2 weeks ago
- YOLOv3-spp COCO (github.com/pjreddie/darknet)
- YOLOv3-spp Open Images v4 (github.com/radekosmulski/yolo_open_images)
- YOLOv3 GIoU COCO (github.com/generalized-iou/g-darknet)
- M2Det 512 COCO (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:
bit.ly/Coffee4Karol

Already received 5 (Updated: 14th April 2019)
Thank you so much!
YOLOv3 object detection vs M2Det  |  COCO vs Open Images v4POLY-YOLO LightOpenVSLAM mapping on Jetson NanoDetectron2: Mask RCNN R50 FPN 3x gnDetectron2: Faster RCNN X101 32x8d FPN 3xIn-Place Activated BatchNorm - Mapillary Vistas - WideResNet38 + DeepLab3 segmentation 8K vizEfficientNet B0 YOLOv3 nightSSD MobileNet v2 Open Images v4Detectron2: Mask RCNN R50 FPN 3xYOLO Object Detection at Night - YOLOv3-spp COCOEdgeTPU object detection - SSD MobileNet V2OpenCV YOLO Object Detection by Adrian Rosebrock (pyImageSearch)
Karol Majek |

YOLOv3 object detection vs M2Det | COCO vs Open Images v4

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER