Dense Extreme Inception Network: Towards
a Robust CNN Model for Edge Detection
Xavier Soria
†
Edgar Riba
†
Angel Sappa
†,‡
† Computer Vision Center - Universitat Autonoma de Barcelona, Barcelona, Spain
‡ Escuela Superior Polit
´
ecnica del Litoral, Guayaquil, Ecuador
{xsoria,eriba,asappa}@cvc.uab.es
Abstract
This paper proposes a Deep Learning based edge de-
tector, which is inspired on both HED (Holistically-Nested
Edge Detection) and Xception networks. The proposed ap-
proach generates thin edge-maps that are plausible for hu-
man eyes; it can be used in any edge detection task without
previous training or fine tuning process. As a second contri-
bution, a large dataset with carefully annotated edges, has
been generated. This dataset has been used for training the
proposed approach as well the state-of-the-art algorithms
for comparisons. Quantitative and qualitative evaluations
have been performed on different benchmarks showing im-
provements with the proposed method when F-measure of
ODS and OIS are considered.
1. Introduction
Edge detection is a recurrent task required for sev-
eral classical computer vision processes (e.g., segmentation
[
39], image recognition [38, 30]), or even in the modern
tasks such as image-to-image translation [
41], photo sketch-
ing [
18] and so on. Moreover, in fields such as medical
image analysis [
27] or remote sensing [16] most of their
heart activities require edge detectors. In spite of the large
amount of work on edge detection, it still remains as an
open problem with space for new contributions.
Since the Sobel operator [
33], many edge detectors have
been proposed [
25] and most of the techniques like Canny
[
5] are still being used nowadays. Recently, in the era
of Deep Learning (DL), Convolutional Neural Netwoks
(CNN) based edge detectors like DeepEdge [
4], HED [36],
RCF [
20], BDCN [14] among others, have been proposed.
These models are capable of predicting an edge-map from a
given image just like the low level based methods [42], with
better performance. The success of these methods is mainly
by the CCNs applied at different scales to a large set of im-
ages together with the training regularization techniques.
Figure 1. The edge-maps predictions from the proposed model in
images acquired from internet.
Most of the aforementioned DL based approaches are
trained on already existing boundary detection or object
segmentation datasets [
22, 31, 24] to detect edges. Even
though most of the images on those datasets are well anno-
tated, there are a few of them that contain missing edges,
which difficult the training, thus the predicted edge-maps
lost some edges in the images (see Fig.
1). In the
current work, those datasets are used just for qualitative
comparisons due to the objective of the current work is
edge detection (not objects’ boundary/contour detection).
The boundary/contour detection tasks, although related and
some times assumed as a synonym task, are different since
just objects’ boundary/contour need to be detected, but not
all edges present in the given image.
This manuscript aims to demonstrate the edge detection
generalization from a DL model. In other words, the model
is capable of being evaluated in other datasets for edge de-
tection without being trained on those sets. To the best of
our knowledge, the unique dataset for edge detection shared
to the community is Multicue Dataset for Boundary Detec-
tion (MDBD—2016) [
23], which although mainly gener-
ated for the boundary detection study, it contains a subset of
1923