drivendataorg / zamba

A Python package for identifying 42 kinds of animals, training custom models, and estimating distance from camera trap videos
https://zamba.drivendata.org/docs/stable/
MIT License
118 stars 27 forks source link

Update expected bounding box in test based on new pillow version #175

Closed ejm714 closed 2 years ago

ejm714 commented 2 years ago

Pillow is used by a number of our packages, including detectron2-densepose, fvcore, pytorchvideo, and torchvision.

The new release of pillow 9.0.0 (https://pypi.org/project/Pillow/#history) causes the detect_image test to fail (example) as the outputted bounding boxes are slightly shifted. I've reproduced this issue locally and can confirm that pillow 8.4.0 yields [0.09714001, 0.04298288, 0.9931407, 1.0082585] whereas 9.0.0 yields [0.09690314, 0.04301501, 0.9931333 , 1.0082883].

This PR updates that test accordingly.

github-actions[bot] commented 2 years ago

🚀 Deployed on https://deploy-preview-175--silly-keller-664934.netlify.app

codecov[bot] commented 2 years ago

Codecov Report

Merging #175 (f78227e) into master (f02284c) will not change coverage. The diff coverage is n/a.

@@          Coverage Diff           @@
##           master    #175   +/-   ##
======================================
  Coverage    85.0%   85.0%           
======================================
  Files          30      30           
  Lines        1851    1851           
======================================
  Hits         1575    1575           
  Misses        276     276