I Stopped Guessing Why My Walk Sucks: Building a Zero-Cloud Street Auditor with Gemma

Description and use

Walkaudit is a local, zero‑cloud Python tool that uses a Gemma vision model to audit street walkability, producing a concise report and actionable civic points.

Problem

Outdoor fitness apps force users to stare at screens, preventing real-time assessment of street walkability.

Outcome

A 350‑meter lane audit yielded an average score of 2.2/5, 50% missing footpath, 81% uneven/broken surface, and identified hazards such as open drains and standing water.

Workflow

• Take photos every 25–30 steps while walking.
• Transfer photos to laptop via USB.
• Run walkaudit locally, which sends images to a local Gemma model via Ollama.
• Model outputs constrained JSON labels for each image.
• Python script aggregates labels, computes a 1–5 score, highlights best/worst spots, and generates civic action points.
• Export audit_report.html and audit.csv.
• Review results on laptop.
• Share findings with community or local authorities.

Evidence

I built walkaudit, a lightweight, zero-cloud Python tool that turns an everyday outdoor walk into an empirical pedestrian walkability audit using a local vision model running via Ollama.

Source

The aggregate audit revealed an average score of 2.2 / 5.0, with 50% of the lane completely missing a footpath and 81% flagging uneven, broken, or unpaved walking surfaces.

Source

Instead of relying on vague impressions like "the road behind our hostel is terrible," walkaudit provides structured civic data:…

Source