# Snapshots (kornia_*.json, cv2_*.json) are regenerated on demand via
# examples/dump_snapshots.sh. They're heavy (~1.6 MB total) and only
# needed during the cv2-parity work — kept out of git on purpose.
#
# CI regression catching uses the committed digests.txt instead: it
# records the SHA256 of each kornia_*.json so any unexpected change
# to find_contours' output is caught by the snapshot_digests integration
# test, without storing the full JSON.
*.json
!.gitignore
