jbaruch/vision-pipeline-foundations
Hygiene patterns for any OpenCV + dlib vision pipeline: camera index probing + macOS init quirks, warmup that verifies real frames, frame-skip policy for expensive inference.
96
93%
Does it follow best practices?
Impact
100%
1.36xAverage score across 6 eval scenarios
Passed
No known issues
Evaluation results
100%
Multi-Worker Face Encoding Pipeline
No VideoCapture in subprocess workers
100%
100%
capture_to_disk pattern
100%
100%
Comment explains single-process camera constraint
100%
100%
No cap.set resolution on DJI Osmo
100%
100%
100%
50%Real-Time Attendance and Mood Monitor
Face recognition skip = EXACTLY 3
0%
100%
Emotion classification skip = EXACTLY 10
0%
100%
Modulo counter gating (not time-based)
100%
100%
Emotion guarded by face presence
100%
100%
Different rates for face vs emotion
100%
100%
100%
30%Conference Demo Camera Script
namedWindow + imshow + waitKey event-loop pump
0%
100%
1.0s warmup sleep (not 0.5s)
100%
100%
Frame mean threshold > 30
100%
100%
No cap.set resolution on DJI Osmo
100%
100%
- Evaluated
- Agent
- Claude
- Model
- Claude Sonnet 4.6