Saving elephants with Raspberry Pi’s & Machine Learning
summary
A field-tested system combines a modified Bushnell camera trap, a microcontroller that wakes a Raspberry Pi, Google AutoML for the classifier and a RockBLOCK satellite modem to send 340‑byte metadata alerts from remote Gabon forests. The prototype skirts slow TensorFlow on Pi by using AutoML, has deployed ~15 units, and successfully alerts rangers to elephants raiding plantations (proving real‑time utility). Next steps: remote software updates and an automated loudspeaker “repeller” tied to detections.
AI-written summary · about
chapters (14)
- 0:00Conference to elephant conservation
- 3:09Gabon’s elephant threats
- 6:52Autoexec.bat to saving elephants
- 11:11Tech as tool not goal
- 14:06Building elephant alert systems
- 17:24ML on Raspberry Pi
- 20:30Imposter syndrome in tech
- 21:58Nature’s human disconnection
- 24:03Testing Gabon camera traps
- 27:57First elephant alert
- 31:22Minister praises tech
- 33:31Coding for purpose
- 35:03Wasted engineering talent
- 38:22Q&A: Poachers and funding
related talks
25mTechnology gone wild
15mThe Internet of Wild Things: Acoustics, AI and Conservation
37mUnstructured Data Processing with a Raspberry Pi AI Kit and Python
17mTechnology: A Changemaker in Environmental Protection
30mProject Sentinel: Home security using Raspberry Pi and Tensorflow
29mSara Hooker & Sean McPherson, Delta Analytics at MLconf Seattle 2017
10mDeep Learning in the Wild: Embedded PyTorch for...
26mMaking an Impact with Duct Tape