Dr. SANJEEVKUMAR ANGADI
@ncerpune.in
ASSISTANT PROFESSOR and COMPUTER SCIENCE AND ENGINEERING
Nutan College of Engineering and Research
EDUCATION
Ph.D. in Computer Science and Engineering
RESEARCH, TEACHING, or OTHER INTERESTS
Computer Vision and Pattern Recognition, Computer Engineering, Computer Science Applications, Artificial Intelligence
Scopus Publications
Scholar Citations
Scholar h-index
Scholar i10-index
Scopus Publications
Neelam Sanjeev Kumar, G. Deepika, V. Goutham, B. Buvaneswari, R. Vijaya Kumar Reddy, Sanjeevkumar Angadi, C. Dhanamjayulu, Ravikumar Chinthaginjala, Faruq Mohammad, and Baseem Khan
Springer Science and Business Media LLC
AbstractA comprehensive examination of human action recognition (HAR) methodologies situated at the convergence of deep learning and computer vision is the subject of this article. We examine the progression from handcrafted feature-based approaches to end-to-end learning, with a particular focus on the significance of large-scale datasets. By classifying research paradigms, such as temporal modelling and spatial features, our proposed taxonomy illuminates the merits and drawbacks of each. We specifically present HARNet, an architecture for Multi-Model Deep Learning that integrates recurrent and convolutional neural networks while utilizing attention mechanisms to improve accuracy and robustness. The VideoMAE v2 method (https://github.com/OpenGVLab/VideoMAEv2) has been utilized as a case study to illustrate practical implementations and obstacles. For researchers and practitioners interested in gaining a comprehensive understanding of the most recent advancements in HAR as they relate to computer vision and deep learning, this survey is an invaluable resource.
M. V. Jagannatha Reddy, J. Somasekar, Kaushalya Thopate, and Sanjeev Kumar Angadi
De Gruyter
C. Senthil Kumar, A. R. Arunarani, Piyush Charan, and Sanjeev Kumar Angadi
De Gruyter
Deepti Chaudhari, Sanjeevkumar Angadi, Saili Sable, Uma Patil, Dipamala Chaudhari, and Kavita Jadhav
IEEE
Sentiment analysis is a vital aspect of understanding public opinion and sentiment towards products and services. This paper presents a sentiment analysis work focused on Zomato restaurant reviews in Bangalore, aiming to classify restaurants into positive, negative, or neutral sentiment categories based on customer reviews. Bi-LSTM and Bi-GRU models are employed to capture contextual information in the sequential data of reviews. Additionally, sentiment analysis techniques, including Word2Vec, VADER, and Sentiment Intensity Analyzer, are integrated to enhance the sentiment classification process. Through rigorous experimentation, the performance of these models and techniques is evaluated. The proposed models demonstrate promising accuracy rates in sentiment classification. By enhancing and expanding the sentiment analysis framework, this paper contributes to a deeper understanding of public sentiment towards Zomato restaurants in Bangalore. The insights derived from this study can facilitate informed decision-making for both restaurant owners and customers, ultimately improving the dining experience and customer satisfaction with accuracy of 98.6%.
Suvarna Nandyal and Sanjeevkumar Angadi
IEEE
one of the active research areas in the field of Computer Vision in today's era is recognizing human activity under video surveillance. To resolve suspicious activity, sensitive and public places such as school, college, jewellery store, railway stations, a temple, bank, etc. can be monitored using video surveillance. It is mind-numbing and time-consuming to track such public areas for a long time. One such area is the Automated Teller Machine (ATM), monitored by a surveillance system. An intelligent monitoring system is proposed to classify real-time based human behaviour and categories them into regular and unusual activities to ensure the safety aspect of ATM and can cause different levels of alarm. This paper proposes a real-time system using the Kalman Filter and the Kanade-Lucas-Tomasi (KLT) Tracking Algorithm to detect and monitor suspicious or non-suspicious human behaviour for ATM video surveillance. On a real-time ATM Surveillance database, experimental results are carried out.
Suvarna Nandyal and Sanjeevkumar Angadi
IEEE
Efficiency of most conventional background subtraction systems used in video surveillance systems depends on the correct choice of a threshold. To prevent this dependency, a new adaptive background modeling method, is proposed in this paper for ATM video monitoring systems, based on the frame averaging method and threshold values. The proposed output of the algorithm was tested on the created ATM data set. The findings of the new approach were compared to those of the traditional Gaussian mixture model. The increased detection efficiency is due to the adaptive threshold introduced in the current background pixel determination process
Sanjeevkumar Angadi and Suvarna Nandyal
IGI Global
Human identification is the most significant topic in the bioinformatics field. Various human gait identification methods are available to identify humans, but detecting the objects based on the human gait is still a challenging task in the video surveillance system. Thus, an effective hybrid Bayesian approach is proposed for identifying the humans. The proposed hybrid Bayesian approach involves two stages as follows: the first stage is the human identification based on the object features, and the second stage is the human identification based on the spatial features. Initially, the videos are fed into the first stage, where the object detection is performed using the Viola Jones algorithm. Once the objects are detected, the feature extraction process is carried out by using a hierarchical skeleton to effectively extract the selective features. The object skeleton provides an effective and intuitive abstraction, which offers object recognition and object matching. The Bayesian network is adapted in the object-based features to identify humans. In the spatial-based human identification stage, only the spatial features are extracted and are passed into the gait-based Bayesian network to identify the humans. Finally, the resulted output is obtained using the fuzzy holoentropy for identifying the humans. The experimentation of the proposed hybrid Bayesian approach is performed using the dataset named UCF-Crime, and the performance is evaluated by considering the average value of the metrics, namely F1-score, precision, and recall which acquired 0.8820, 0.8770, and 0.9203, respectively.