Docker image for visual regression tests
After extensive testing, I found DSSIM & ꟻLIP to be the most accurate metrics for comparing images (at least for my specific needs).
By using both tools I can improve accuracy & minimize errors. Since there's no Docker image with ꟻLIP, I decided to provide my own.
NOTE ꟻLIP source files are taken from Nvidia's website https://research.nvidia.com/sites/default/files/pubs/2020-07_FLIP/flip-cpp.zip.
I am not the creator of this fantastic image evaluator.
flip <reference.{jpg|png}> <test.{jpg|png}> [options]
Options:
-help # Show this text
-v <[0-2}> # Verbosity: 0 = silent, 1 = pooled values (default), >1 = verbose
-heatmap <heatmap.png> # Generate heatmap image
-nomagma # Grayscale (FLIP values) instead of Magma heatmap (with "-heatmap")
-histogram <filename> # Pooling output (<filename>.csv and <filename>.py files generated)
-log # Log10 on y-axis in the Python pooling histogram (with "-histogram").
0 - images are the same
1 (?) - images are completely different
You know how to use it :) But anyway:
dssim file-original.png file-modified.png
Content type
Image
Digest
Size
12.5 MB
Last updated
almost 6 years ago
docker pull ozonep/visual-diff:v3