Semantic Line Segment Detector

Sem-LSD: A Learning-based Semantic Line Segment Detector,

This repo contains code to run Sem-LSD, a learning-based and light-weighted detector for a new line-shaped image representation, Semantic Line Segment (Sem-LS).

This project is built based on CenterNet codebase.

Installation

Please refer to INSTALL.md for installation instructions.

Data preparation

We use training and validating data in COCO format. Please refer to DATA.md for instructions.

Training & Inference & Evaluation

To train / inference and perform evaluation using our proposed ACL metric, please refer to RUNNING.md for details.

Pretrained models are also provided.