Identifying Vehicle Exterior Color by Image Processing and Deep Learning
Date
2022-04-19
Authors
Abniki, Somayeh
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The vehicle’s color is one of the factors considered in car purchasing.
Hence, color extraction and identification from online vehicle images play an
important role in the vehicle e-commerce marketplace. In this paper, we present
a vehicle color identification methodology. Image processing techniques are
employed to construct feature vectors, which are then used as input to deep neural
networks to classify a vehicle’s color into 14 classes. Local relative entropy is
utilized as a measure of image segmentation to select the region of interest.
Experiments are performed on an image dataset provided by an automobile ecommerce
operator. Our implementation results are evaluated and discussed.
Description
Keywords
entropy, local relative entropy, color recognition, DNN