Annotators can now press Auto Segment to replace the current mask with
an automatic river segmentation based on dense optical flow magnitude
and frame brightness. The result is pushed onto the undo stack, so it
can be refined or reverted like any other mask operation.
Parameters (norm_squared_threshold, gaussian_kernel, brightness_range)
live in a separate config/optical_flow_config.yaml; the button is only
enabled when optical_flow_config_file is set in config.yaml.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- config.py constants -> config/config.yaml (user-editable, git-ignored)
- Questions and defaults now defined in the YAML, including per-question defaults
- ClipSelector no longer scans the data dir; reads a user-provided clips.txt instead
- Removed --daily / --time / --skip-existing-day args
- video_loader now samples frames evenly across the full clip
- pyyaml added as a dependency
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
