search

vasgaowei / pytorch_MELM

The pytorch implementation of the Min-Entropy Latent Model for Weakly Supervised Object Detection
104 stars 19 forks source link

Did you use the ‘gt_bbox’ annotation in your model? #8

Closed magemoumou closed 5 years ago

magemoumou commented 5 years ago

I looked all through your code. Then I found that during the generation of the 'roidb', you used the 'gt_bbox' annotation saved in the file: pytorch_MELM/data/VOCdevkit2007/VOC2007/Annotations. But the original model tends be a "Weakly Supervised Detection Model" which just using the image label annotation without the 'gt_bbox' I believe.

vasgaowei commented 5 years ago

No, I don't use the "gt_box". You can check my code carefully. I don't change the 'roidb' generation code, but during the forward propagation, I don't use the 'gt_box'.

magemoumou commented 5 years ago

OK, I finally run through your code. Maybe I am supposed to transform my dataset to pascal voc format or not change your code. Thanks.