# Project export: PathSense

This document was generated by HackStack to give an AI agent context about a hackathon project. Sections are labeled with their provenance; content marked as truncated was cut to keep this document small.

## Project metadata

- Hackathon: Cal Hacks 12.0
- Tagline: Blind-first navigation with an AI safety net
- Devpost: https://devpost.com/software/pathsense-t16vmo
- GitHub: not linked
- Video: https://www.youtube.com/embed/v2mqXE42WM4?enablejsapi=1&hl=en_US&rel=0&start=&version=3&wmode=transparent
- Team: contributor stats unavailable

## Devpost submission (written by the team)

### Inspiration

A white cane is great, but it can’t “see” fast-moving hazards. We wanted a low-cost, phone-only co-pilot that narrates the scene and nudges safe movement—without sending video to the cloud.

### What it does

PathSense runs on-device object detection and immediately speaks the object first (“person ahead”, “chair ahead”) and then a navigation cue (“slight left”, “stop”, “caution”). It streams lightweight event logs to a Fetch.ai Agentverse safety agent that: Detects danger states (stuck, repeated near-collisions, or device silence) Sends SMS alerts to a trusted contact Archives telemetry to ASI One for privacy-scoped insights that only the caregiver can access

### How we built it

Android (Kotlin, CameraX, Foreground Service, Jetpack Compose) YOLOv8n → TFLite INT8 (320×320, NMS output [300×6]) via NNAPI/XNNPACK A planner that maps detections → tokens (STOP / CAUTION / SLIGHT LEFT / RIGHT) with smoothing and dead-zones TTS + haptics for clear, low-latency cues Agentverse client (OkHttp) to stream events; Agent triggers SMS and stores logs in ASI One with caregiver-only credentials

### Challenges we ran into

Quantized model scaling & confidence thresholds across devices Label/index mismatches and coordinate normalization Reducing “chatter”: debouncing speech, stabilizing cues, handling orientation Balancing latency vs. battery in a persistent camera service

### Accomplishments we're proud of

True on-device guidance with clear object-then-action narration End-to-end safety oversight via Agentverse + ASI One with strict access control Robust mobile pipeline that survives screen-off and poor connectivity

### What we learned

Edge AI is as much systems work as model work: logging, smoothing, permissions, and UX (the right words at the right time) matter as much as mAP.

### What's next

Stairs/curb depth cues Finer path planning Wearable audio integration Multilingual voices Caregiver dashboards on ASI One Structured pilot with blind users to iterate on safety and trust

## README (from the GitHub repository)

No README available.

## Detected evidence (automated analysis)

No repository was indexed for this project. Claimed technologies below could not be checked against code.
- Kotlin (language) — claimed on Devpost, not found in the code
- Python (language) — claimed on Devpost, not found in the code
- TensorFlow (technology) — claimed on Devpost, not found in the code

## Codebase structure

No repository index available.

## Key source files

No repository index available; no source files included.