Ananda Chowdhury

Ananda Shankar Chowdhury is a Professor and former Head in the department of Electronics and Telecommunication Engineering at Jadavpur University, Kolkata, India, where he leads the Imaging, Vision and Pattern Recognition group. He earned his Ph.D. in Computer Science from the University of Georgia, USA in July 2007. He then worked as a post-doctoral fellow in the department of Radiology and Imaging Sciences at the National Institutes of Health, USA during the period August 2007 to December 2008. His research interests span Computer Vision, Pattern Recognition, Biomedical Image/Signal Processing, and Multimedia Analysis. Dr. Chowdhury has published more than one hundred papers in leading international journals and conferences. He has also published two books from Morgan Kaufmann, Elsevier and Advances in Computer Vision and Pattern Recognition (ACVPR) Series of Springer. He is a senior member of IEEE, an International Association for Pattern Recognition (IAPR) TC-member of Graph based Representations in Pattern Recognition, and a life member of the Indian Unit of Pattern Recognition and Artificial Intelligence (IUPRAI). He has held multiple invited academic visits to universities in Germany, France, Norway, Italy, The Netherlands, Singapore and Brazil. He has delivered more than fifty invited lectures and tutorials in India and abroad, in addition to a plenary talk in a conference organized by the Calcutta Mathematical Society, and a short course at Indian Institute of Science. At present, he serves as an Associate Editor for IEEE Transactions on Image Processing, and as an Area Editor for Pattern Recognition Letters. His Erdӧs Number is 2.

Website: https://sites.google.com/site/anandachowdhury/

Contact: as.chowdhury@jadavpuruniversity.in

Abstracts

Computer Vision for Detection and Prediction of Cancer - A Tale of Many Seasons

Abstract - In this talk, I will take you with me to an eventful journey through different seasons of computer vision with an aim of detection and prediction of cancer (restricted to lungs and brain). In the first season, which is considered to be “Autumn”, I will address the problem of segmentation of lung nodules, a potential precursor to lung cancer, from CT images, using only level sets devoid of any form of learning. We will wait for a relatively unproductive “Winter” for some new form of machine learning, viz. deep learning, to flourish. Then, we will proceed to our third season, “Spring”. Here, many flowers blossom, which can be equivalently thought of as a beautiful synergism of deep learning and classical computer vision. Segmentation of brain tumors in 3D from MRI data using a combination of deep learning and graph cuts will be discussed here [2]. The journey will come to an end with the fourth and final “Rainy” season, where deep learning rains or reigns. Here, we will first demonstrate how attributes driven generative adversarial networks can synthesize and classify various types of lung nodules [3]. As a second problem of this season, it will be shown how five important genetic markers responsible for glioma, the most common form of malignant brain tumor, can be predicted using deep learning with a loss function having roots in probability and graph theory [4].

References

[1] R. Roy, P. Banerjee, A.S. Chowdhury, A Level Set based Unified Framework for Pulmonary Nodule

Segmentation, IEEE Signal Processing Letters 27 (2020), 1465-1469.

[2] A. De, M. Tewari, E. Grisan, A.S. Chowdhury, A Deep Graph Cut Model for 3D Brain Tumor Segmentation, Proc. Forty-fourth International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), Glasgow, Scotland, UK (2022), 2105 - 2109.

[3] R. Roy, S. Mazumdar, A.S. Chowdhury, ADGAN: Attributes Driven Generative Adversarial Network for Synthesis and Multi-class Classification of Pulmonary Nodules, IEEE Transactions on Neural Networks and Learning Systems 35(2) (2024), 2484 - 2495.

[4] A. De, A. Santra, M. Tiwari, A.S. Chowdhury, Predicting Genetic Markers for Brain Tumors Using a Composite Loss, IEEE/ACM Transactions on Computational Biology and Bioinformatics (2025) (in press)

(doi: 10.1109/TCBBIO.2025.3593318)

Note: The uploaded video is an excerpt of a recording of a guest lecture I have given at Arctic University of Norway in October 2022 with a similar title and content. The excerpt shows how the topic has been introduced.

Link: https://drive.google.com/file/d/1tzGVgqkqqemp_K8qbWoP7-0Ec0GEfUQb/view?usp=drive_link

Recognize Actions in Egocentric Video, Retrieve Images at Large in Limited Annotated Data Scenario

Abstract - In this talk, I will show how tasks like action recognition from egocentric video and retrieval of images at large, which could easily become further challenging due to limited availability of annotated data, are handled with efficacy using weakly supervised/unsupervised models. We will mainly focus on three problems in this context. In the first problem, we will address recognition and localization of single as well multiple actions in egocentric videos within a weakly supervised setup. Our solution will be based on random walks applied to a sparsely connected video representation graph of superpixels in a deep feature space. Some theoretical results on the density and connectedness of the above graph will be shown [1-2]. In the second problem, we will demonstrate how fine-grained actions, and composite activities can be detected in smart home environments with attention networks and Bayesian inference on knowledge graphs amid scarcity of labelled training data [3]. Finally, in the third problem we will present a patch-based unsupervised deep image representational model, termed as Bag of Variational Deep Embedded Visual Words for facilitating large-scale image retrieval together with bipartite graph based object part matching. The loss function used in this solution will explicitly be shown as convex [4].

References

[1] A. Sahu, A. S. Chowdhury, Together Recognizing, Localizing and Summarizing Actions in Egocentric Videos, IEEE Transactions on Image Processing 30 (2021), 4330 - 4340.

[2] A.S. Chowdhury, A. Sahu, Graph Based Multimedia Analysis, Morgan Kaufmann, Elsevier, Cambridge, MA, USA, ISBN: 978-0-443-21495-0. (2024)

[3] S. T. Kumbhare, A.S. Chowdhury: Activity Recognition in Smart Homes with Knowledge Graph and Attention-Guided Learning, Ninth IAPR International Conference on Computer Vision and Image Processing; Chennai, India (2024), J. Kakarla (eds), Springer CCIS, Vol. 2476, 454 - 467.

[4] A. Mukherjee, J. Sil, A. Sahu, A.S. Chowdhury: Patch based Unsupervised Deep Learning for Retrieval and Part Matching in Large Image Datasets, IEEE Transactions on Big Data (2025) (in press) (doi: 10.1109/TBDATA.2025.3594237)

Real-world Video Surveillance – From Person Re-identification to Moving Object Detection in Event Camera 

Abstract - In this talk, I will address some key problems in real-world surveillance. Three different problems will be discussed, two on person re-identification [1-3], and the third on moving object detection in event camera [4]. Person re-identification focusses on matching individuals across non-overlapping camera views. In the first problem, we will illustrate how deep dictionary learning can provide a lightweight solution for partial and occluded person re-identification [2]. For the second problem of scale invariant person re-identification, deep metric learning based solutions will be presented [3]. Event cameras are essentially neuromorphic vision sensors that can capture a stream of asynchronous events in contrast to the traditional RGB cameras which acquire images at a fixed-frame rate. As the third problem, we will focus on moving object detection from event camera using a graph spectral clustering approach [4].

References

[1] A. Sikdar, A.S. Chowdhury: Efficient Deep Learning for Key Challenges in Person Re-identification, in Pattern Recognition and Computer Vision in the New AI Era, C.H. Chen (eds.), World Scientific Publishing, Singapore (2025), 399 - 426.

[2] A. Sikdar, A.S. Chowdhury: Lightweight Learning for Partial and Occluded Person Re-identification, IEEE Transactions on Artificial Intelligence 6(5) (2024) 3245 - 3256.

[3] A. Sikdar, A.S. Chowdhury: Scale-Invariant Batch-Adaptive Residual Learning for Person Re-identification, Pattern Recognition Letters 129 (2020), 279 - 286.

[4] A. Mondal, Shashant R., J.H. Giraldo, T. Bouwmans, A.S. Chowdhury: Moving Object Detection for Event-based Vision using Graph Spectral Clustering, First ICCV Workshop on Graph Signal Processing meets Computer Vision (GSP-CV), Montreal, Canada (2021), 876 - 884.