A new baseline for edge detection: Make Encoder-Decoder great again

TL;DR

This paper addresses the trade-off between high edge-detection accuracy and computational efficiency by proposing NBED, a vanilla encoder–decoder that decouples location and semantic feature extraction via a bilateral, lightweight Location Feature Encoder and a CNN–Transformer hybrid Semantic Feature Encoder. A Cascaded Feature Fusion Decoder progressively refines location features with semantic information, ensuring the final edge map is derived from purified location features, thereby suppressing texture noise and localization errors. NBED is trained with a simple annotator-robust loss and achieves state-of-the-art or competitive results on BSDS500, NYUDv2, and BIPED while maintaining low GFLOPs, suggesting that high-quality feature extraction is more critical than complex training pipelines. The approach offers a practical, scalable baseline for edge detection with broad applicability and potential extension to other pixel-wise tasks.

Abstract

The performance of deep learning based edge detector has far exceeded that of humans, but the huge computational cost and complex training strategy hinder its further development and application. In this paper, we eliminate these complexities with a vanilla encoder-decoder based detector. Firstly, we design a bilateral encoder to decouple the extraction process of location features and semantic features. Since the location branch no longer provides cues for the semantic branch, the richness of features can be further compressed, which is the key to make our model more compact. We propose a cascaded feature fusion decoder, where the location features are progressively refined by semantic features. The refined location features are the only basis for generating the edge map. The coarse original location features and semantic features are avoided from direct contact with the final result. So the noise in the location features and the location error in the semantic features can be suppressed in the generated edge map. The proposed New Baseline for Edge Detection (NBED) achieves superior performance consistently across multiple edge detection benchmarks, even compared with those methods with huge computational cost and complex training strategy. The ODS of NBED on BSDS500 is 0.838, achieving state-of-the-art performance. Our study shows that what really matters in the current edge detection is high-quality features, and we can make the encoder-decoder based detector great again even without complex training strategies and huge computational cost. The code is available at https://github.com/Li-yachuan/NBED.

Figures (4)

• Figure 1: (a) Overall framework of NBED. SeparConvs and Attentions contain multiple SeparConv and Attention modules, respectively. n$\times$ down means n $\times$ downsampling. $E_i$: Feature size at stage i of the encoder, $D_i$: Feature size at stage i of the decoder. (b) Meta block. SeparConv and Attention in (a) are composed of Meta block and different Operators.
• Figure 2: Qualitative comparisons on BSDS dataset with previous state-of-the-arts.
• Figure 3: Qualitative comparison on BIPED dataset with the previous state-of-the-art.
• Figure 4: Qualitative comparison of different encoders