Session profile
- Audience: teams operating proofreading or segmentation-QC workflows.
- Duration: 80 minutes lecture + 15 minutes correction-log exercise.
- Output: prioritized error triage plan and QC dashboard spec.
Slide-by-slide lecture plan
- Slide 1 (2 min): Title and QC framing
- Slide 2 (5 min): Why proofreading determines scientific validity
- Slide 3 (6 min): Error taxonomy
- merge, split, boundary, identity errors.
- Slide 4 (6 min): Human-machine division of labor
- Slide 5 (6 min): Correction protocol (local to global consistency)
- Slide 6 (6 min): Metrics deep dive I
- VI, edge precision/recall.
- Slide 7 (6 min): Metrics deep dive II
- ERL and synapse-centric F1.
- Slide 8 (7 min): Prioritization by downstream impact
- Slide 9 (7 min): Worked correction case I
- Slide 10 (7 min): Worked correction case II
- Slide 11 (5 min): Release gating and audit logs
- Slide 12 (5 min): Team operations and reviewer calibration
- Slide 13 (5 min): Failure modes and anti-patterns
- Slide 14 (7 min): Activity + debrief
- write a correction rationale and metric update.
- Primary shortlist:
course/units/figures/08-segmentation-and-proofreading-selected-v1.md.
- Include at least one before/after correction sequence and one metric dashboard panel.
Speaker notes (expert-level)
- Emphasize high-impact-first correction strategy.
- Pair aggregate metrics with targeted biological sanity checks.
Assessment and artifacts
- Deliverable: correction log with before/after rationale and metric deltas.
- Rubric dimensions: error classification, correction quality, and reproducibility.
Connections
Slide source file
- Marp draft source:
course/decks/marp/08-segmentation-and-proofreading.marp.md
- Batch render helper:
./scripts/render_marp.sh
