Logo detection using weakly supervised saliency map
- PDF / 6,567,058 Bytes
- 25 Pages / 439.642 x 666.49 pts Page_size
- 86 Downloads / 225 Views
Logo detection using weakly supervised saliency map Gautam Kumar1 · Prateek Keserwani1 · Partha Pratim Roy1 · Debi Prosad Dogra2 Received: 17 July 2018 / Revised: 11 August 2020 / Accepted: 2 September 2020 / © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract Box level annotation of a large number of logo images for training purpose of typical deep learning architecture is highly challenging. Thus, a method that can detect the logo with the help of training to remove box-level annotations can be helpful. In this paper, we present a method of logo detection that utilizes weakly supervised learning of Convolutional Neural Network (CNN) to generate a deep saliency map. The saliency map is generated from the back-propagated response of the CNN trained with the classification task. The saliency map produces responses for the regions of logos. GrabCut segmentation method has been applied then to obtain the bounding box corresponding to the logo class predicted by the CNN for a given image. AlexNet, CaffeNet, and VGGNet deep architectures has been finetuned for the classification purpose. The framework is further utilized for detection through a back-propagated saliency map. The performance of the proposed methodology has been validated on the FlickrLogos-32 logo benchmark dataset. The proposed method outperforms the state-of-the-art baseline fully supervised methods with mean average precision (mAP) of 75.83%. Keywords Convolutional neural network · Saliency map · GrabCut · Logo detection · Weakly supervised
Gautam Kumar
[email protected] Prateek Keserwani [email protected] Partha Pratim Roy [email protected] Debi Prosad Dogra [email protected] 1
Department of CSE, Indian Institute of Technology Roorkee, Roorkee, India
2
School of Electrical Sciences, IIT Bhubaneswar, Bhubaneswar, India
Multimedia Tools and Applications
1 Introduction The online visual data repository is increasing rapidly due to the fast expansions of social networking sites. This massive increment in visual data needs robust and efficient solution for object detection and recognition [18, 20, 50] to support content-based retrieval. A Logo is a special class of the object, which is extremely important to find an item’s belonging to a specific brand. Generally, the design of the logo consists of symbols and texts. It often emphasizes the product or company name and shows specific meaning for the object. The logo represents the brand and represents the quality and services of the product. Logos are the key feature for readers to describe the ownership of visual documents. Logo detection and recognition systems have gained importance in the research community because of its wide range of applications. Some of its uses are namely, vehicle logo recognition, traffic management, product brand management in social media. Logo detection tries to identify the locations of the logos within an image. Detected regions can further be assigned to the specific class of the logo, known as logo recognition. It can b
Data Loading...
