jbaruch/vision-pipeline-foundations-kotlin
Hygiene for JavaCV + DJL vision pipelines on Kotlin/JVM: camera discovery and probing, frame-skip policy for heavy inference, downscale-before-detection. Replaces the Python jbaruch/vision-pipeline-foundations tile.
94
93%
Does it follow best practices?
Impact
99%
1.86xAverage score across 3 eval scenarios
Passed
No known issues
Evaluation results
100%
44%Webcam Discovery Utility for macOS Development Workstation
Camera probe and selection utility
Probe range 0-5
100%
100%
500 ms warm-up
0%
100%
Brightness threshold 10.0
0%
100%
opencv_core.mean for brightness
75%
100%
Safe grabber.stop()
100%
100%
CAM env var override
33%
100%
imageWidth/imageHeight set before start
0%
100%
AutoCloseable Camera
100%
100%
Probe result data structure
100%
100%
No hardcoded index 0
100%
100%
100%
46%Optimizing a Real-Time Face Recognition Pipeline for Live Events
Frame-skip inference loop with downscale
Face detection every 3rd frame
25%
100%
Emotion every 30th frame
25%
100%
Grab on every iteration
100%
100%
Persisted overlay variable
100%
100%
4x downscale before detection
0%
100%
Bounding boxes scaled up 4x
0%
100%
Recognition on full-resolution frame
100%
100%
No delay between grabs
100%
100%
Detection Hz diagnostic
100%
100%
No full-res detection
0%
100%
99%
49%Reactive Vision Pipeline Using Kotlin Coroutines
Flow-based reactive vision pipeline
sample(100ms) for detection
0%
100%
sample(1s) for emotion
0%
100%
Continuous frame producer
100%
100%
4x downscale before detection
0%
100%
Bounding boxes scaled up 4x
0%
100%
Full-resolution recognition crop
100%
90%
Persisted overlay in preview
100%
100%
No delay in producer
100%
100%
Separate detection and emotion flows
100%
100%
- Evaluated
- Agent
- Claude
- Model
- Claude Sonnet 4.6