First,I found the jde.1088x608.uncertainty.pt is at epoch 60,so I started a training process for 60 epochs。All settings are original in the code includes the dataset setting。At last,I get a result with mAP=0.7888 while jde.1088x608.uncertainty.pt can attach mAP = 0.8209. Because of the gap in mAP, it also gets a lower MOTA metric。Dose anyone also take the experiment as mine?
First,I found the jde.1088x608.uncertainty.pt is at epoch 60,so I started a training process for 60 epochs。All settings are original in the code includes the dataset setting。At last,I get a result with mAP=0.7888 while jde.1088x608.uncertainty.pt can attach mAP = 0.8209. Because of the gap in mAP, it also gets a lower MOTA metric。Dose anyone also take the experiment as mine?