# cardwing/Codes-for-Lane-Detection

Learning Lightweight Lane Detection CNNs by Self Attention Distillation (ICCV 2019)

Repository: https://github.com/cardwing/Codes-for-Lane-Detection
Canonical: https://ross.abutalabs.com/products/codes-for-lane-detection
Language: Lua
License: MIT
License Family: permissive
Topics: lane-detection, deep-learning, lua, tensorflow, cnn, pytorch
Last push: 2022-06-21T21:48:30+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 2882, "days_push": 1534, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1075, forks 333 (observed 2026-08-28T04:03:29.320817+00:00)

## What it is
Reference implementations of lightweight lane detection CNNs, including the ENet-SAD model from the ICCV 2019 paper 'Learning Lightweight Lane Detection CNNs by Self Attention Distillation' plus a TensorFlow reimplementation of SCNN. Models are provided in Torch/Lua and PyTorch with training and testing code for the TuSimple, CULane, and BDD100K datasets.

## Use cases
- detect lane markings in dashcam or driving footage
- train a lightweight lane detection model on CULane
- reproduce self attention distillation results from the ICCV 2019 paper
- benchmark lane detection on TuSimple and BDD100K
- run SCNN in TensorFlow for traffic scene understanding
- deploy a fast low-parameter lane segmentation network for autonomous driving research

## When to choose
- you need a lightweight, fast lane detection model with published benchmarks
- you want to reproduce or build on the SAD or SCNN papers
- you are doing autonomous driving perception research on TuSimple, CULane, or BDD100K

## When to avoid
- you need a production-ready, actively maintained lane detection pipeline
- you want a plug-and-play detector without touching research code
- you need support for modern framework versions (code targets TensorFlow 1.x and older PyTorch)
- your task is general semantic segmentation rather than lane detection

## Facets
- artifact type: library
- maturity: maintenance
- function: deep-learning, computer-vision, machine-learning, image-processing
- domain: computer-vision, autonomous-vehicles, deep-learning, machine-learning
- platform: python
- tags: lane-detection, self-attention-distillation, knowledge-distillation, semantic-segmentation, cnn, pytorch, tensorflow, torch, autonomous-driving, research-code, enet, scnn, erfnet, tusimple, culane, bdd100k, linux, gpu

## Member repositories
- cardwing/Codes-for-Lane-Detection (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:29.320817+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T06:53:25.951383+00:00, confidence not recorded.
  - readme: https://github.com/cardwing/Codes-for-Lane-Detection (fetched 2026-08-28T04:03:29.320817+00:00, sha 6eebf501bad3)
- Data as of 2026-08-30T08:39:29.467469+00:00.