Abstract
Image restoration and recognition are important computer vision tasks representing an inherent part of autonomous systems. These two tasks are often implemented in a sequential manner, in which the restoration process is followed by a recognition. In contrast, this paper proposes a joint framework that simultaneously performs both tasks within a shared deep neural network architecture. This joint framework integrates the restoration and recognition tasks by incorporating: (i) common layers, (ii) restoration layers and (iii) classification layers. The total loss function combines the restoration and classification losses. The proposed joint framework, based on capsules, provides an efficient solution that can cope with challenges due to noise, image rotations and occlusions. The developed framework has been validated and evaluated on a public vehicle logo dataset under various degradation conditions, including Gaussian noise, rotation and occlusion. The results show that the joint framework improves the accuracy compared with the single task networks.
Highlights
Image recognition is an important field, especially for autonomous systems
Previous works [2,21] show that the capsule network approach achieves better results on image recognition than conventional convolutional network networks (CNNs)
We have developed three implementations and have tested them over different image degradation including noise, rotation and occlusion
Summary
Image recognition is an important field, especially for autonomous systems. The field witnessed new opportunities and became very popular with the development of convolutional network networks (CNNs) in 2012 [15]. Compressed sensing algorithms incoprorating second-order derivatives in efficient ways are developed in [6], approaches based on conjugate priors in [7] and on supporting machines in [19] In such approaches, image features are extracted at pixel level or regional level and embedded in the recognition process [32]. The algorithms reported in [10,14,27,30] achieve state-of-the-art performance on a single image with super-resolution [9]. These methods are only applicable to particular types of image degradation. The approaches for learning restoration are promising as they do not need models nor assumptions about the nature of the degradation [18]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
