Detecting Honeybee Swarms Using the Integration of OpenCV, Pandas, AI, & PyTorch
summary
Camera captures a still image every 30 seconds to count bees at the entrance, images uploaded to S3 and preprocessed with OpenCV, counts obtained via Marvin.AI or a local YOLO V8 model trained with iNaturalist data in PyTorch, aggregated into a Pandas rolling window 10‑minute moving average, alerts sent via AWS SNS within ~10 minutes of a detected swarm; hardware uses solar‑powered Raspberry Pi and long‑range Ubiquiti links; results so far are inconclusive (hive did not swarm).
AI-written summary · about
chapters (28)
- 0:00Honeybee swarm detection intro
- 1:06Why honeybees swarm in spring
- 3:56Hive structure and swarm process
- 6:56Swarm detection challenges
- 7:17Hive monitoring hardware setup
- 9:27Solar power for monitoring
- 12:20Raspberry Pi camera setup
- 13:46Remote monitoring internet access
- 15:13System monitoring importance
- 16:18Python AI integration
- 18:27Image capture storage strategy
- 19:51Uploading images to AWS S3
- 20:35OpenCV preprocessing techniques
- 22:25Contrast Limited Adaptive Histogram Equalization
- 23:27Marvin AI for bee counting
- 24:10YOLO v8 training with PyTorch
- 25:33Image annotation challenges
- 26:16AI bee counting models
- 27:41Swarms vs bearding distinction
- 28:48Moving average swarm analysis
- 29:51Pandas rolling Z-score analysis
- 30:55Results and improvements
- 31:15Conclusion and acknowledgments
- 31:44Q&A: Swarming timing
- 32:29Q&A: Annotation tools
- 33:47Q&A: Wildlife monitoring extension
- 34:51Q&A: Acoustic mite monitoring
- 35:13Q&A: Alternative CV models
related talks
24mWhat's the Buzz with Machine Learning
01h 23mTutorial: A Practical Introduction To Machine Learning
20mFINDING BAD BEES IN THE HIVE
49mMonitoring global events with AI-driven satellite swarms
23mBirding with Python and Machine Learning
57mDrones & AI - What's all the buzz about ?
13mLoRaWAN-Driven Beehive Monitoring with Students
25mTechnology gone wild