Pushing the boundaries of boundary detection using deep learning - Archive ouverte HAL Access content directly
Conference Papers Year : 2016

Pushing the boundaries of boundary detection using deep learning

(1, 2)
1
2

Abstract

In this work we show that adapting Deep Convolutional Neural Network training to the task of boundary detection can result in substantial improvements over the current state-of-the-art in boundary detection. Our contributions consist firstly in combining a careful design of the loss for boundary detection training, a multi-resolution architecture and training with external data to improve the detection accuracy of the current state of the art. When measured on the standard Berkeley Segmentation Dataset, we improve theopti-mal dataset scale F-measure from 0.780 to 0.808 - while human performance is at 0.803. We further improve performance to 0.813 by combining deep learning with grouping, integrating the Normalized Cuts technique within a deep network. We also examine the potential of our boundary detector in conjunction with the task of semantic segmentation and demonstrate clear improvements over state-of-the-art systems. Our detector is fully integrated in the popular Caffe framework and processes a 320x420 image in less than a second.
Not file

Dates and versions

hal-02432711 , version 1 (08-01-2020)

Identifiers

  • HAL Id : hal-02432711 , version 1

Cite

Iasonas Kokkinos. Pushing the boundaries of boundary detection using deep learning. 4th International Conference on Learning Representations, ICLR 2016, May 2016, San Juan, Puerto Rico. ⟨hal-02432711⟩
81 View
0 Download

Share

Gmail Facebook Twitter LinkedIn More