Zippy is a simple, lightweight, and quick image annotation tool for creation of image classification groundtruth. It is built using feh - the fenomenal image viewer and cataloguer.

The decisions to take are assign label 1, 2 or skip until all images are exhausted or program is quit manually.

• install feh
• copy zippy somewhere into the PATH

Run Zippy as:

zippy path_to_images

Zippy will traverse the path recursively, randomize the image order.

full paths and assigned labels

/data/eagle.jpg,1
/data/dog.jpg,2
/data/colibri.png,1
/data/albatros.jpg,1
/data/camel.jpg,2
/data/deer.png,2
…

full paths to skipped files

/data/trout.png
/data/sardine.jpg

• Enter - skips the image and adds to skipped.txt
• 1 - assigns label 1 and adds the record to labels.txt
• 2 - assigns label 2 and adds the record to labels.txt
• q - quit zippy
• see feh's manual for the rest

• resume
• input labels as arguments
• active training loop
• allow label correction
• …