Learning to Decode Contextual Information for Efficient Contour Detection

ABSTRACTContour detection plays an important role in both academic research and real-world applications. As the basic building block of many applications, its accuracy and efficiency highly influence the subsequent stages. In this work, we propose a novel lightweight system for contour detection that achieves state-of-the-art performance while keeps ultra-slim model size. The proposed method is built on an efficient encoder in a bottom-up/top-down fashion. Specially, we propose a novel decoder that compresses side features from an encoder and effectively decodes compact contextual information for high-accurate boundary localization. Besides, we propose a novel loss function that is able to assist a model to produce crisp object boundaries. We conduct extensive experiments to demonstrate the effectiveness of the proposed system on the widely adopted benchmarks BSDS500 and Multi-Cue. The results show that our system achieves the same best performance, yet only consumes 3.3% computational cost (16.45GFlops VS. 499.15GFlops) and 2.35% model size (1.94M VS. 82.43M) of the SOTA detector RCF-ResNet101. In the meantime, our method outperforms a large portion of the recent top edge detectors by a clear margin.