# Foresight House AI

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Diana Olynick](https://cerebralvalley.ai/u/DianaOlynick)
- **GitHub:** https://github.com/foresighthouseai/wildlife-road-alert
- **Demo video:** https://drive.google.com/file/d/1NQetLdq6UJco9HMAfV-0sfNskiuHIDaF/view?usp=drive_link
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/359

Wildlife Road Alert turns the dashcams already on the road into an early-warning network for wildlife collisions. In Manitoba alone there are 47 wildlife collisions a day, and at highway speed a driver who sees a deer in the headlights has about one second to react: too late to stop.

The system watches dashcam video with SAM 3.1, finding both the animal and the road in every frame from a single word, with no training. Our rules turn the animal's size and growth into distance and seconds to impact, and raise a green, amber or red warning, including a "more may follow" hold for herds. Muse Spark double-checks the species. Every detection becomes a shared sighting on a map, so the car that sees the deer gains seconds and the vehicles behind it gain minutes: in our demo, a truck 4 km back is warned 2.6 minutes before it arrives. Drivers can also report animals by voice (Muse Voice Transcribe), Muse Wonder writes a nightly briefing for road operators, and operator-approved sightings are formatted as a Waze feed so drivers without a camera are warned too.

We measured it: on 282 road photos labelled by people, SAM 3.1 found 487 of 508 deer (96%) with zero training, including 31 of 32 far-away deer and 21 of 22 at night. On a real night dashcam clip it spotted the deer about 44 m ahead and went red 0.6 s later.

Today it runs on recorded footage. Next: real time, in the vehicle, starting with fleets that already drive these roads with cameras.

## More from Meta Global AI Developer Hackathon

- [Kareem](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/356)
- [0xInan](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/357)
- [Wisora](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/358)
- [BlinkFlow](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/360)
- [PulseGuard](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/361)
- [ARIQ Water](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/362)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/359. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt