Redesigned YOLOv2:

For training, we train redesigned YOLOv2 with 150 epochs on COCO.
For data augmentation, we only use the large scale jitter (LSJ), no Mosaic or Mixup augmentation.
For optimizer, we use SGD with momentum 0.937, weight decay 0.0005 and base lr 0.01.
For learning rate scheduler, we use linear decay scheduler.

Model	Backbone	Batch	Scale	AP^val 0.5:0.95	AP^val 0.5	FLOPs ^(G)	Params ^(M)	Weight
YOLOv2	DarkNet-19	1xb16	640	32.7	50.9	53.9	30.9	ckpt