---
library_name: ultralytics
pipeline_tag: image-segmentation
tags:
- onnx
- yolov8
- image-segmentation
- object-detection
- civic-tech
- traffic-safety
- bike-lanes
- crosswalks
---

# Reported Bike Lane and Crosswalk Detector

This repository contains the Reported YOLOv8 segmentation ONNX export used for complaint pre-categorization in civic traffic reporting workflows.

The model artifact is:

- `reported-v13-optimized.onnx`

## Model Details

- Format: ONNX
- Architecture: YOLOv8s segmentation
- Input size used by the local postprocessor: 512 x 512
- Source metadata: `Ultralytics YOLOv8s-seg model trained on /content/Reported-13/data.yaml`
- SHA-256: `64ff4e935369defa74e30fcc4c19474f37f418cf6b6c76a1560d7e9570442879`

## Classes

The embedded class metadata lists 13 classes:

| Index | Label |
|---:|---|
| 0 | blocked bike lane |
| 1 | blocked crosswalk |
| 2 | bullbars |
| 3 | ct |
| 4 | fl |
| 5 | md |
| 6 | medallion |
| 7 | nc |
| 8 | nj |
| 9 | ny |
| 10 | nypd |
| 11 | pa |
| 12 | tlc |

## Intended Use

Use this model to detect and segment street-safety complaint cues, especially blocked bike lanes and blocked crosswalks, in images or video frames. The repository includes `labels.json` for downstream apps that need the class index mapping.

## Limitations

This model should be used as an assistive signal, not as the sole source of truth for enforcement or other high-impact decisions. Users should review detections and confirm report details before submitting them to a city or agency.

## Postprocessing Notes

The Reported app repository includes a YOLOv8 segmentation postprocessor for this ONNX export. It treats class `0` as `blocked bike lane`, applies confidence and IoU thresholds, and decodes mask coefficients from the model outputs.