Add source_model column to hard_negatives table with migration. New
get_hard_negatives() returns full rows, delete_hard_negatives_by_ids()
for bulk deletion. get_training_data() gains use_hard_negatives param.
TrainDialog has "Use hard negatives" checkbox. Scan panel passes current
model name when marking negatives.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add scan_timestamp column to scan_results. save_scan_results now inserts
with a timestamp and prunes versions beyond max_versions (default 5).
get_scan_results returns only the latest version by default, with optional
scan_timestamp parameter for loading specific versions. New get_scan_versions
method returns available versions for a (file, profile, model) tuple.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Ghost folders (scan-export-only) no longer appear in training dropdowns.
Also filters out 0-clip folders from get_training_stats.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Fix test_utils.py importing build_annotation_json_path from main
instead of core.annotations (all 59 tests pass now)
- Fix get_training_data double-counting clips at same start_time
in both positive and soft sets — subtract positive from soft
- Add cancel_flag to train_classifier so training can be interrupted
between videos (TrainWorker passes self as cancel_flag)
- Remove orphaned core/export.py (was for deleted server API)
- Remove stale Dockerfile and docker-compose.yml (referenced server)
- Clean up leftover server/__pycache__ and client/ build artifacts
- Add torch to requirements.txt (was only mentioned in comments)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Remove legacy distance-mode scanning (build_profile, _similarity, etc.)
and hand-crafted intensity features — pipeline is now embedding-only
- Integrate Microsoft BEATs as embedding option alongside wav2vec2/HuBERT
- Add TrainDialog with positive class selector, model picker, video dir
fallback, and live training stats
- Add TrainWorker QThread with cancel support and proper lifecycle cleanup
- Add source_path column to DB for robust source video tracking
- Add get_export_folders/get_training_data/get_training_stats to DB
- Wire source_path in all export DB writes (_on_clip_done, _on_auto_clip_done)
- Cancel scan/train workers in closeEvent to prevent use-after-free crashes
- Add setup_env.sh supporting both conda and python venv (CUDA 12.8)
- Update requirements.txt with all actual dependencies
- Update 8cut_train.py with --positive flag for new DB-driven training
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Root cause of poor discrimination: MFCC[0] (energy) dominated the
feature vector, making cosine similarity see all audio as similar.
Changes:
- Skip MFCC[0], use 12 coefficients instead of 20
- Add delta MFCCs for temporal dynamics
- Add 7-band spectral contrast for tonal vs noise quality
- Switch from cosine similarity to euclidean-distance-based score
- Pre-compute STFT once for whole file (10-20x faster)
- Vectorized sliding window via cumulative sums (no Python loop)
- Lower sample rate 22050→16000 Hz (faster, no quality loss)
- 62-dim feature vector (was 40-dim mean+std of raw MFCCs)
- Default threshold 0.05 (new similarity scale)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Mean-only vectors were too similar across different audio segments,
causing everything to match even at threshold 0.99. Adding std
captures temporal dynamics and makes the similarity scores much
more spread out.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Adds a pure function that returns the latest keyframe at or before a
given time (with tolerance), replacing the inline lookup pattern that
appears multiple times in main.py. Includes 6 tests covering empty
list, before-first, exact match, between, after-last, and tolerance.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Lock button (G key) freezes export cursor, timeline scrubs playback only
- In lock mode, clicking crop bar sets a keyframe at current playback time
- Orange diamonds on timeline show keyframe positions
- Export resolves per-clip crop center from nearest preceding keyframe
- Crop bar/overlay updates while scrubbing to preview effective crop
- Unlocking clears all keyframes
- Replace fuzzy filename matching with exact match to prevent marker bleed
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Each profile has its own set of timeline markers, so the same video
can be cut with different settings (e.g. landscape vs portrait) without
markers interfering. Profile selector in the top bar, persisted in
QSettings, stored per-row in the DB.
- Add `profile` column to ProcessedDB schema (migrates existing rows
to 'default')
- Scope find_similar, get_markers, _get_markers_for by profile
- Add get_profiles() for populating the combo dropdown
- Thread profile through _DBWorker, _after_load, _refresh_markers,
_on_clip_done, dropEvent, _on_open_files
- Editable profile QComboBox in top bar, refreshed after each export
- 5 new tests for profile isolation and backward compatibility (54 total)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Style QComboBox/QSpinBox/QDoubleSpinBox/QCheckBox in dark theme
- Delete and overwrite now operate on the full clip group, not just one sub-clip
- Add get_group/delete_group to ProcessedDB with tests
- Restructure control rows: transport+actions / annotation+path / encoding
- Add "Open Files" button to queue panel (replaces drag-drop-only)
- Playlist right-click to remove items
- Compact time display (1:23.4 / 5:00.0), window title shows filename
- Short side: QLineEdit → QSpinBox with validation
- Tooltips with keyboard shortcuts on all interactive widgets
- Fix arrow hint direction, remove stale mask comment
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Dead code — masking is handled externally via ComfyUI. Removes
SetupWorker, MaskWorker, SettingsDialog, build_mask_output_dir,
the mask UI row, Settings button, and associated test cases.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Each batch export creates a subfolder named after the group (e.g.
clip_001/) containing all sub-clips and their audio files. Keeps
the top-level export folder clean.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Each export generates clip_NNN_0, clip_NNN_1, clip_NNN_2 offset by the
spread value (2–8s, default 3s). Preview plays the full span covered by
all three clips. Marker click still overwrites a single clip.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Each exported clip writes an entry to <folder>/dataset.json containing
its relative path, sound label, and fps. Re-exporting to the same path
updates the existing entry (upsert). Empty labels are skipped.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Adds build_audio_extract_command and runs it in ExportWorker after the
frame sequence completes. Audio written to <sequence_dir>.wav (lossless
pcm_s16le). Extraction failure (no audio stream) is silently ignored.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Ethanfel