search

ryouchinsa / Rectlabel-support

RectLabel is an offline image annotation tool for object detection and segmentation.
https://rectlabel.com
507 stars 73 forks source link

Is it possible to add option to make the image frames to 1:1 square size when convert video to image frames #171

Closed caichunjian520 closed 4 years ago

caichunjian520 commented 4 years ago

It is a good choice to take a video and convert it to image frames. If i take video with iPhone, the images are rectangle that is not fit for square size usage. Is it possible to add an option to crop all images to square, i think crop bottom part is easy for us to keep the object always in images when take video.

ryouchinsa commented 4 years ago

Thanks for writing the issue.

that is not fit for square size usage.

Usually training modules crop training images according to the input size of the neural network. Let us know the detailed use-case why images should be cropped to square on RectLabel.

ryouchinsa commented 4 years ago

Currently If you need our support to fix this problem, please let us know.