Showing papers by "Shyamanta M. Hazarika published in 2021"
TL;DR: This paper proposes a novel approach combining diagrammatic reasoning with qualitative spatial and temporal reasoning techniques to visualize and perceive spatio-temporal relations among objects in a video.
3 citations
TL;DR: This work revisits the implicit design choices in the popular vector of locally aggregated descriptors (VLAD), which aggregates the residuals of descriptors to the codewords and proposes a division of codebook feature space using a novel fine-grained quantization strategy.
Abstract: We are interested in the encoding of local descriptors of an image (e.g. SIFT) to design a compact representation vector and thereby address scalable image retrieval. We revisit the implicit design choices in the popular vector of locally aggregated descriptors (VLAD), which aggregates the residuals of descriptors to the codewords. VLAD’s use of a coarse codebook and first-order descriptor statistics in residual computation results in less discriminative residuals. To address this problem, we propose a division of codebook feature space using a novel fine-grained quantization strategy. After quantization, we embed the resulting residuals with high-order statistics of descriptor distribution. Experiments on three challenging image retrieval datasets (INRIA Holidays, UKBench, Oxford 5k) confirm the improved discriminative power of our novel encoding method called FhVLAD. We observe superior accuracy to baseline and competitive performance to state-of-the-art techniques with a limited increase in dimension.
2 citations
DOI•
20 Sep 2021TL;DR: In this article, a framework for EEG-based emotion recognition using Multi Layer Perceptron (MLP) was proposed, which quantifies the emotions in terms of valence-arousal scale and MLP is used for classification.
Abstract: Emotion Recognition is an important problem within Affective Computing and Human Computer Interaction. In recent years, various machine learning models have provided significant progress in the field of emotion recognition. This paper proposes a framework for EEG-based emotion recognition using Multi Layer Perceptron (MLP). Power Spectral Density features were used for quantifying the emotions in terms of valence-arousal scale and MLP is used for classification. Genetic algorithm is used to optimize the architecture of MLP. The proposed model identifies a. two classes of emotions viz. Low/High Valence with an average accuracy of 91.10% and Low/High Arousal with an average accuracy of 91.02%, b. four classes of emotions viz. High Valence-Low Arousal (HVLA), High Valence-High Arousal (HVHA), Low Valence-Low Arousal (LVLA) and Low Valence-High Arousal (HVHA) with 83.52% accuracy. The reported results are better compared to existing results in the literature.
2 citations