Deep Learning model for compressing UI screenshots when a related reference screenshot exists. Compares conventional codecs with learned no-reference and reference-conditioned codecs on 24k WebSight-derived pairs, with a local demo and a mode selector that falls back to no-reference.
deep-learning pytorch image-compression visual-regression-testing websight learned-image-compression compressai neural-codec ui-screenshots
-
Updated
Oct 31, 2026 - Python