I think in ssd_vgg_preprocessing.py, the function preprocess_for_train is making dataset enhancement or making more training data,do we need make the same change when change the image size ,then change the boxes?
dst_image = tf_image.resize_image()
I only see here resize the image, but didnt change the boxes location
Hi @kingrain213 this issue has been discussed in #237
maybe you can reference "pascalvoc_to_tfrecords.py" in folder of "datasets"
you can see boxes location has been rescaled between 0~1
I think in ssd_vgg_preprocessing.py, the function preprocess_for_train is making dataset enhancement or making more training data,do we need make the same change when change the image size ,then change the boxes?
dst_image = tf_image.resize_image() I only see here resize the image, but didnt change the boxes location