I waited for the 174th epoch, but mAP wasn't improved, so I didn't think that this issue is related to training time.
I also tried to change hyper- parameters such as learning rate, batch size,... and augmentation such as common size = 640 and scale parameter in augmentation.py, but mAP didn't be improved. Has anyone any solution for this problem?
I tried to train efficientdet -d1 model on VOC dataset and got low mAP. mAP: aeroplane: 0.37774546979552437 bicycle: 0.2717609619189093 bird: 0.07373408655526852 boat: 0.11174968193861216 bottle: 0.03085867327429357 bus: 0.3259476835112142 car: 0.4270658513682306 cat: 0.1865584138729217 chair: 0.039742404670251544 cow: 0.08904280802274998 diningtable: 0.1571207438493799 dog: 0.11370612854293503 horse: 0.3255647249082447 motorbike: 0.32683360884097407 person: 0.2828047853052909 pottedplant: 0.006887167196655821 sheep: 0.13986957637436875 sofa: 0.15581529346760506 train: 0.36374159709309817 tvmonitor: 0.2485077007219078 avg mAP: 0.20275286806142181
I waited for the 174th epoch, but mAP wasn't improved, so I didn't think that this issue is related to training time. I also tried to change hyper- parameters such as learning rate, batch size,... and augmentation such as common size = 640 and scale parameter in augmentation.py, but mAP didn't be improved. Has anyone any solution for this problem?