Exploring Real-time Computer Vision Using ActionCable
summary
Concrete claim: a Rails app integrates OpenCV with ActionCable & Sidekiq and Python services to analyze cannabis leaf area and bud-site area in real time, producing masks, counts, per-site size, and alerts. Implementation details include HSV calibration, scikit-image measurements, edge options like Raspberry Pi + Movidius VPU, cloud options via Thumbor on Google Cloud Run, and storage in Postgres/Redis with Puma/Sidekiq. Demonstrated results: tiling cut drone processing from minutes to <1s and an example growth of 27,918 pixels = ~27 cm² in 5 days.
AI-written summary · about
chapters (15)
- 0:00Intro
- 1:08Tesla Computer Vision
- 3:01Dairy Barn Computer Vision
- 4:34Computer Vision Experiences
- 7:28Cannabis Computer Vision
- 9:55Implementation
- 11:54Leaf Area Analysis
- 12:59Bud Area Analysis
- 14:19Architecture Overview
- 16:06Edge Compute on Raspberry Pi
- 19:33Direct File Uploads
- 22:08WebRTC Signaling with ActionCable
- 23:24Does it Scale?
- 27:16Overcoming Manual Calibration
- 29:00Off The Shelf Solutions
related talks
26mDetecting and classifying object images using ruby
30mWhen performance matters: running a real-time computer vision pipeline
37mUnstructured Data Processing with a Raspberry Pi AI Kit and Python
24mBeyond the Frame: Using Computer Vision to Unlock the Potential of Video Data
36m"Lights, camera, AI!" - Automated sports videography
40mComputer Vision Using Ruby and libJIT
20mScaling Edge AI with Elixir and Nerves
26mUse computer vision at the edge to improve operations with AWS Panorama