237 research outputs found
Discriminative body part interaction mining for mid-level action representation and classification
No full textn this paper, we propose a novel mid-level feature representation for the recognition of actions in videos. This descriptor proves to posses relevant discriminative power when used in a generic action recognition pipeline. It is well known that mid-level feature descriptors learnt using class-oriented information are potentially more distinctive than the low-level features extracted in a bottom-up unsupervised fashion. In this regard, we introduce the notion of concepts, a mid-level feature representation capable of tracking the dynamics of motion salient regions over consecutive frames in a video sequence. Our feature representation is based on the idea of region correspondence over consecutive frames and we make use of an unsupervised iterative bipartite graph matching algorithm to extract representative visual concepts from action videos. The progression of such salient regions, which are also consistent in appearance, are henceforth represented as chain graphs. Finally, we adopt an intuitive time-series pooling strategy to extract discriminant features from the chains, which are then used in a dictionary learning based classification framework. Given the high variability of the movements of different human body parts in separate actions, the extracted conceptual descriptors are proved to capture the different dynamic characteristics by exclusively encoding the interaction parts associated to the chains. Further, we use such descriptors in a semi-supervised, clustering-based zero-shot action recognition setting, showing good performance and without resorting to costly attribute annotation. We validate the proposed framework on four public datasets namely KTH, UCF-101, HOHA and HMDB-51, reporting increased (and comparable in some cases) classification accuracies with respect to the state of the art. © 2018 Elsevier Inc
Single Image Super-Resolution for Optical Satellite Scenes Using Deep Deconvolutional Network
No full textIn this paper, we deal with the problem of super-resolution (SR) imaging and propose a deep deconvolutional network based model for the same. In principle, the SR problem considers the construction of the high-resolution (HR) version of a scene given a number of so-called low-level image instances of the respective scene. Moreover, if there is a single low-resolution (LR) image available, the problem becomes even difficult and ill-posed. We deal with such a scenario and show how the popular deconvolutional network can effectively reconstruct the HR image by learning the functional mapping at the patch level. We evaluate the proposed model on a number of optical remote sensing (RS) images obtained from the UC-Merced dataset. Experimental results suggest that the proposed model consistently outperforms the existing deep and shallow models for single image SR for the RS images
Semantic Guided Deep Unsupervised Image Segmentation
No full textImage segmentation is an important step in many image processing tasks. Inspired by the success of deep learning techniques in image processing tasks, a number of deep supervised image segmentation algorithms have been proposed. However, availability of sufficient labeled training data is not plausible in many application domains. Some application domains are even constrained by the shortage of unlabeled data. Considering such scenarios, we propose a semantic guided unsupervised Convolutional Neural Network (CNN) based approach for image segmentation that does not need any labeled training data and can work on single image input. It uses a pre-trained network to extract mid-level deep features that capture the semantics of the input image. Extracted deep features are further fed to trainable convolutional layers. Segmentation labels are obtained using argmax classification of the final layer and further spatial refinement. Obtained segmentation labels and the weights of the trainable convolutional layers are jointly optimized in iterations in a mechanism that the deep network learns to assign spatially neighboring pixels and pixels of similar feature to the same label. After training, the input image is processed through the same network to obtain the labels that are further refined by a segment score based refinement mechanism. Experimental results show that our method obtains satisfactory results inspite of being unsupervised
A Simplified Framework for Zero-shot Cross-Modal Sketch Data Retrieval
No full textWe deal with the problem of zero-shot cross-modal imageretrieval involving color and sketch images through a noveldeep representation learning technique. The problem of asketch to image retrieval and vice-versa is of practical im-portance, and a trained model in this respect is expectedto generalize beyond the training classes, e.g., the zero-shot learning scenario. Nonetheless, considering the dras-tic distributions-gap between both the modalities, a fea-ture alignment is necessary to learn a shared feature spacewhere retrieval can efficiently be carried out. Additionally,it should also be guaranteed that the shared space is se-mantically meaningful to aid in the zero-shot retrieval task.The very few existing techniques for zero-shot sketch-RGBimage retrieval extend the deep generative models for learn-ing the embedding space; however, training a typical GANlike model for multi-modal image data may be non-trivialat times. To this end, we propose a multi-stream encoder-decoder model that simultaneously ensures improved map-ping between the RGB and sketch image spaces and highdiscrimination in the shared semantics-driven encoded fea-ture space. Further, it is guaranteed that the class topologyof the original semantic space is preserved in the encodedfeature space, which subsequently reduces the model biastowards the training classes. Experimental results obtainedon the benchmark Sketchy and TU-Berlin datasets estab-lish the efficacy of our model as we outperform the existingstate-of-the-art techniques by a considerable margin
Structural Adjustment, Global Trade and the New Political Economy of Development
No full textBiplab Dasgupta. Structural Adjustment Global Trade and the New Political Economy of Development. New Delhi. Sage Publications. 1998. Price Indian Rupees 450 (hardback). The author has written a very topical book the relevance of which cannot be understated. At the core of the book the author discusses the concept of the new political economy of development which forms the theoretical underpinnings that lie behind the structural adjustment/ stabilisation programmes of the international financial institutions such as the World Bank and the International Monetary Fund. Biplab Dasgupta has very concisely and succinctly analysed the new political economy of development which has, as its centre-piece, a blind faith in the operation of free-market forces. This can be traced back to the Reagan and Thatcher years, which saw a shift away from interventionist policies to allowing the markets to decide
Biplab Majumdar and His Poetry with Special Attention to Cosmic Convergence
No full textBiplab Majumdar is the author of nearly 100 books of poetry, prose, rhyme, translation, novels and short stories. His works are published both in Bengali and English. The contents of this latest volume, Cosmic Convergence, are divided into two parts: Part-A covers the year from January to December and Part-B contains 12 poems on a variety of subjects. The poems are followed by 3 pages of selected comments on Biplab Majumdar’s by a variety of eminent authors. This volume makes possible an assessment of the scope and stature of Majumdar’s work. These poems-often witty and beautiful- are an achievement, a testament to Majumdar’s ongoing power to engage us in his vision. They confirm Majumdar’s reputation as one of India’s finest poets. From evocations of the daily wonders of life to explorations of spirituality, feelings and sensibilities. His celebration of idiom and understanding of the modern mind may help us to understand ourselves
Binding Characteristics of Anticancer Drug Doxorubicin with Two-Dimensional Graphene and Graphene Oxide : Insights from Density Functional Theory Calculations and Fluorescence Spectroscopy
No full textThere has been a perpetual interest in identifying suitable nano-carriers for drug delivery. In this regard, graphene-based two-dimensional materials have been proposed and demonstrated as drug carriers. In this paper, we have investigated the adsorption characteristics of a widely used anticancer drug, doxorubicin (DOX), on graphene (G) and graphene oxide (GO) by density functional theory calculations and fluorescence and X-ray photoelectron spectroscopies. From the calculated structural and electronic properties, we have concluded that G is a better binder of DOX compared to GO, which is also supported by our fluorescence measurements. The binding of DOX to G is mainly based on strong pi-pi stacking interactions. Consistent with this result, we also found that the sp(2) regions of GO interact with DOX stronger than the sp(3) regions attached with the functional groups; the binding is characterized by pi-pi and hydrogen-bonding interactions, respectively.</p
A two-coil mutual inductance technique to study the conductivity of metal and measurement of the superconducting transient temperature
No full textInformation Seeking Behavior of Research Scholars of Vidyasagar University, West Bengal
No full textThe main objective of the study is to investigate the information seeking behavior of the research scholars of Vidyasagar University (VU), West Bengal. Besides, the study also intended to identify their information needs and awareness regarding the library services rendered by the central library of the university. Required data was collected from 100 researchers of the university through a structured questionnaire. Findings indicate that guidance in the use of library resources and services is necessary to help researchers to meet some of their information requirements. Most of the researchers avail library weekly (30%) and monthly(45%) and spend maximum of 0-2 hours with the main intention for preparing research, study journals , up-to-date knowledge etc. Most of the researchers depend on VU central library (27%) for information seeking and also collect information from other sources
- …
