Open ithuestad opened 4 years ago
+1
+1
+1
if you copy select labels and boxes from the previous frame into the current one, then current frame isn't useful for your task, specialize in machine learning. Your model can't learning any more if you have two or more image with the same position and same coordinates of bounding box. So i guess you can remove your similarity images and just keep one.
@AnhPC03 I'd argue that strongly depends on the specific cases. I currently have many cases where we label pictures with multiple classes, where some classes stay static in a series of images while the others move. If training data is sparse, there is no good reason to not use these images for training, even if only class detection of the moving classes benefits. However you still have to label also the static classes in these images. So for this cases (and I have them with over 300 objects per image) this feature would be an immense time saver EDIT: Typo
Along the lines of reddts, I have a scenario with some static objects and different light conditions so this feature would be helpful.
+1
Hey guys, if i'm understanding well this feature was already added in this PR https://github.com/tzutalin/labelImg/pull/627
Thank you campos537, you saved my day =)
press: Ctrl + v to copy the labels of the previous image
It would be nice to copy select labels and boxes from the previous frame into the current one, possibly with the press of a key. Having this capability would speed up the labeling process for video frames.