Apply the same transformation to MixedItem

fastai
1

Note this can not be addressed by ‘tfm_y=True’ since here x is a mixed item that has two parts: an image and a bounding box. Usually, bboxes are labels, so we could simply set tfm_y to be true.

by image.apply_tfms(tfms[0], **kwargs); bboxes.apply_tfms(tfms[0], **kwargs);, it gives two different transforms (although the same type of transforms)

image
image.jpg1219×819 254 KB

2

simply setting do_resolve=False

[image.apply_tfms(tfm, **kwargs).show(ax=ax, y=bboxes.apply_tfms(tfm, do_resolve=False, **kwargs)) for i,ax in enumerate(plt.subplots(
        rows,cols,figsize=(width,height))[1].flatten())]