Replies: 1 comment 2 replies
-
Hi, have you checked out the |
Beta Was this translation helpful? Give feedback.
2 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hi, have you checked out the |
Beta Was this translation helpful? Give feedback.
-
Hi, I'm using tensorcom to serve batched
(images,labels)
pairs. The labels are are padded bboxes with class ids, so their size is(N, max_boxes, 5)
(where the 5 are x,y,w,h,c).For example, when
max_boxes=16
,labels[0]
may look like:What's the most efficient way to input these pairs into an image/bbox augmentation pipeline? Do the augmentations support padded bbox batches? I thought that if not, converting the tensor to a list of unpadded tensors might not be efficient.
Beta Was this translation helpful? Give feedback.
All reactions