Open wuminjie12 opened 3 years ago
I have trained my datasets with pretrained d2. While i found the val loss going up i stopped it.The test result is as below. There are two classes. The handwrite data and the seal. Only the seal can be detected.
The mAP result doesn't explicitly show the AP of each class, so it's possible that mAP seems fine while the realistic result sucks.
I have trained my datasets with pretrained d2. While i found the val loss going up i stopped it.The test result is as below. There are two classes. The handwrite data and the seal. Only the seal can be detected.