Tiny and Efficient Model for the Edge Detection Generalization

TL;DR

This paper introduces TEED, a lightweight and efficient edge detection model with only 58K parameters, designed for simplicity, speed, and generalization, outperforming larger models in training speed and ease of use.

Contribution

The paper presents TEED, a novel small CNN for edge detection that is fast to train, easy to optimize, and generalizes well across different datasets.

Findings

TEED has only 58K parameters, less than 0.2% of SOTA models.

Training on BIPED takes less than 30 minutes, with rapid convergence.

TEED produces high-quality, crisp edge maps.

Abstract

Most high-level computer vision tasks rely on low-level image operations as their initial processes. Operations such as edge detection, image enhancement, and super-resolution, provide the foundations for higher level image analysis. In this work we address the edge detection considering three main objectives: simplicity, efficiency, and generalization since current state-of-the-art (SOTA) edge detection models are increased in complexity for better accuracy. To achieve this, we present Tiny and Efficient Edge Detector (TEED), a light convolutional neural network with only $58K$ parameters, less than $0.2$% of the state-of-the-art models. Training on the BIPED dataset takes $less than 30 minutes$, with each epoch requiring $less than 5 minutes$. Our proposed model is easy to train and it quickly converges within very first few epochs, while the predicted edge-maps are crisp and of high quality. Additionally, we propose a new dataset to test the generalization of edge detection, which comprises samples from popular images used in edge detection and image segmentation. The source code is available in https://github.com/xavysp/TEED.