monocongo / cvdata

Tools for creating and manipulating computer vision datasets
MIT License
17 stars 9 forks source link

Fix edge case with annotation resizing #162

Open tehtea opened 3 years ago

tehtea commented 3 years ago

How to replicate original bug on Colab:

%%bash

wget http://host.robots.ox.ac.uk/pascal/VOC/voc2012/VOCtrainval_11-May-2012.tar
tar xf VOCtrainval_11-May-2012.tar
unconverted_train_image_folder = "VOCdevkit/VOC2012/JPEGImages/"
unconverted_train_annot_folder = "VOCdevkit/VOC2012/Annotations/"

train_image_folder = 'data/train/image/'
train_annot_folder = 'data/train/annotation/'
!cvdata_resize --input_images $unconverted_train_image_folder \
    --input_annotations $unconverted_train_annot_folder \
    --output_images $train_image_folder \
    --output_annotations $train_annot_folder \
    --width $IMAGE_W --height $IMAGE_H --format pascal