Over the last decade, Convolutional Neural Network (CNN) models have been highly successful in solving complex vision problems. However, these deep models are perceived as "black box" methods considering the lack of understanding of their internal functioning. There has been a significant recent interest in developing explainable deep learning models, and this paper is an effort in this direction. Building on a recently proposed method called Grad-CAM, we propose a generalized method called Grad-CAM++ that can provide better visual explanations of CNN model predictions, in terms of better object localization as well as explaining occurrences of multiple object instances in a single image, when compared to state-of-the-art. We provide a mathematical derivation for the proposed method, which uses a weighted combination of the positive partial derivatives of the last convolutional layer feature maps with respect to a specific class score as weights to generate a visual explanation for the corresponding class label. Our extensive experiments and evaluations, both subjective and objective, on standard datasets showed that Grad-CAM++ provides promising human-interpretable visual explanations for a given CNN architecture across multiple tasks including classification, image caption generation and 3D action recognition; as well as in new settings such as knowledge distillation.

No abstract

No abstract

Abstract:In this communication, a novel compact high gain composite right/left-handed (CRLH) based leaky-wave antenna (LWA) is presented at Ku band. Half-mode substrate integrated waveguide (HMSIW) incorporating with suitably oriented Complementary Quad Spiral Resonator (CQSR) is used to achieve a CRLH LWA. The uni cell is realized by a CQSR in such a way that orientation of spirals exhibit higher leakage loss having minimum cross coupling between them. The antenna is capable to scan backward to forward along with broadside direction in visible space. The proposed configuration is just length of 4.85λ 0 which can scan within the frequency range of 13.5-17.8 GHz having beam scanning range of 86• (-66• to 20 • ) and maximum gain of 16 dBi. The simulated reflection coefficient of the proposed antenna is below -10 dB throughout the working frequency range with a side-lobe level of below -10 dB. The designed prototype is much compact in nature having high gain, fair scanning range, good cross-polarization level along with simpler design methodology and tuning capability to enhance the gain as well as radiation efficiency maintaining fixed size. The proposed antenna could be a promising candidate in Ku-band applications like Fixed Satellite Services (FSS) and Broadcast Satellite Services (BSS) etc.