Department of Computer Science & Engineering

&ldquo;IIT Jodhpur is committed to impart quality education to facilitate the creation of qualified, competent and responsible human resources to meet the emerging technological challenges of the world. It is working towards creating an equitable learning platform that is positive, accessible and effective in the face of the modern societal and scientific advancements.

Flexibility in curriculum, in all its academic programs, encompassing elements of teaching, research, creativity, innovation and entrepreneurship is the most unique feature of IIT Jodhpur. The Department of Science and Technology (DST), under the National Mission on Cyber Physical Systems, has identified IIT Jodhpur as a Technology &amp; Innovation Hub in the area of Augmented Reality and Virtual Reality. The institute has also been nominated as the coordinator for the Jodhpur City and Knowledge Cluster by the Office of Principal Scientific Adviser, Government of India.

The sprawling 852-acre IIT Jodhpur campus is a majestic tribute to the deep-rooted diverse cultural heritage of the region. The campus is an architectural masterpiece, which is slated to become an international exemplar of sustainability with net-zero energy, and advanced water and waste management strategies.&rdquo;

IIT Jodhpur

A trained radiologist may learn the visual presentation of a new disease by looking at a few relevant image examples in research articles. However, training a machine learning model in such a manner is an arduous task not only due to the small number of labeled training images but also for the low resolution of such images. We design a few-shot learning method that can diagnose new diseases from chest x-rays utilizing only a few relevant labeled x-ray images from the published literature. Our method uses prior knowledge about other diseases for feature extraction from x-rays of new diseases. We formulate a classifier that is initially trained with a few labeled feature vectors corresponding to low-resolution images from the PubMed Central. The classifier is subsequently re-trained using unlabeled feature vectors corresponding to high-resolution x-ray images. Experiments on publicly available datasets show the superiority of the proposed method to several state-of-the-art few-shot learning techniques for chest x-ray diagnosis. © 2021 IEEE.

World University root node

Proceedings - International Symposium on Biomedical Imaging

Learning few-shot chest X-ray diagnosis using images from the published scientific literature

The copyright holder for this preprint is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY 4.0 International license. Purpose: To advance the usage of CXRs as a viable solution for efficient COVID-19 diagnostics by providing large-scale annotations of the abnormalities in frontal CXRs in BIMCV-COVID19+ database and to provide a robust evaluation mechanism to facilitate its usage. Materials and Methods: We provide the abnormality annotations in frontal CXRs by creating bounding boxes. The frontal CXRs are a part of the existing BIMCV-COVID19+ database. We also define four different protocols for robust evaluation of semantic segmentation and classification algorithms. Finally, we benchmark the defined protocols and report the results using popular deep learning models as a part of this study. Results: For semantic segmentation, Mask-RCNN performs the best among all the models with a DICE score of 0.43 ± 0.01. For classification, we observe that MobileNetv2 yields the best results for 2-class and 3-class classification. We also observe that deep models report a lower performance for classifying other classes apart from the COVID class. Conclusion: By making the annotated data and protocols available to the scientific community, we aim to advance the usage of CXRs as a viable solution for efficient COVID-19 diagnostics. This large-scale data will be useful for ML algorithms and can be used for learning radiological patterns observed in COVID-19 patients. Further, the protocols will facilitate ML practitioners for unified large-scale evaluation of their algorithms.

medRxiv

A novel abnormality annotation database for COVID-19 affected frontal lung X-rays

Deep neural networks are generally trained using large amounts of data to achieve state-of-the-art accuracy in many possible computer vision and image analysis applications ranging from object recognition to natural language processing. It is also claimed that these networks can memorize the data which can be extracted from the network parameters such as weights and gradient information. The adversarial vulnerability of the deep networks is usually evaluated on the unseen test set of the databases. If the network is memorizing the data, then the small perturbation in the training image data should not drastically change its performance. Based on this assumption , we first evaluate the robustness of deep neural networks on small perturbations added in the training images used for learning the parameters of the network. It is observed that, even if the network has seen the images it is still vulnerable to these small perturbations. Further, we propose a novel data augmentation technique to increase the robustness of deep neural networks to such perturbations.

2021 IEEE International Conference on Image Processing (ICIP)

Intelligent and Adaptive Mixup Technique for Adversarial Robustness

IEEE Transactions on Pattern Analysis and Machine Intelligence

DeriveNet for (Very) Low Resolution Image Classification

Deaths and injuries are common in road accidents, violence and natural disaster. In accidents and natural disasters scenarios, one of the tasks of responders is to retrieve the identity of the victims to reunite families or ensure proper identification of deceased persons. Apart from this, the identification of unidentified dead bodies due to violence and accidents is crucial for the police investigation. In the absence of identification cards, different forensic techniques such as DNA profiling and dental profiling may be used for identification. In this research, we present face recognition as a fast and viable approach for recognizing individuals with injuries. Face, which can be captured easily, is one of the most commonly used and widely accepted biometric modalities. However, face recognition is challenging in the presence of injuries as facial injuries change the appearance and geometric properties of the face due to swelling, bruises, blood clots and accidental cuts. These changes introduce large intra-class variations among the same subject and small inter-class separability among different subjects. To address the challenge, we propose a novel Subclass Injured Face Identification (SCIFI) loss which is used in learning feature representation agnostic to injury variations. Additionally, an extended Injured Face (IF-V2) database of 150 subjects is presented to evaluate the performance of face recognition models. Multiple experiments and comparisons are performed to showcase the efficacy of the proposed SCIFI loss based face recognition.

IEEE Transactions on Biometrics, Behavior and Identity Science

Recognizing Injured Faces via SCIFI Loss

Proceedings of the AAAI Conference on Artificial Intelligence 35 (18), 15745 {\ldots}

Role of Optimizer on Network Fine-tuning for Adversarial Robustness ( Supplementary )

Fulltext

Kidney International Reports

Pos-578 Short Term Outcomes of Tunneled Cuffed Catheters- a Single Center Experience

The Bitcoin-Next Generation (NG) is a new blockchain protocol that is designed to scale the efficiency in terms of throughput and delay. Bitcoin-NG is a Byzantine fault-tolerant blockchain protocol robust to extreme churn and shares the same trust model as Bitcoin. Researchers are trying to extend their applications by integrating them with existing technologies, such as the Internet of Things (IoT). However, due to high computational power and high consensus delay, current blockchain protocols, such as Bitcoin, are not suitable for working with lightweight IoT devices. Moreover, they are not energy-efficient as the mining of blocks requires dedicated mining machines that consume electricity. However, protocols, such as Bitcoin-NG, are designed to reduce the consensus delay, but they still need high computational power and high energy. This article proposes a new mechanism for blockchain to reduce the consensus delay, reduce energy consumption, and increase the throughput by introducing a new leader election scheme in the blockchain. © 2014 IEEE.

IEEE Transactions on Computational Social Systems

Toward Next Generation of Blockchain Using Improvized Bitcoin-NG

Classical vertex subset problems demanding connectivity are of the following form: given an input graph G on n vertices and an integer k, find a set S of at most k vertices that satisfies a property and G[S] is connected. In this paper, we initiate a systematic study of such problems under a specific connectivity constraint, from the viewpoint of Kernelization (Parameterized) Complexity. The specific form that we study does not demand that G[S] is connected but rather G[S] has a closed walk containing all the vertices in S. In particular, we study Closed Walk-Subgraph Vertex Cover (CW-SVC, in short), where given a graph G, a set X⊆ E(G), and an integer k; the goal is to find a set of vertices S that hits all the edges in X and can be traversed by a closed walk of length at most k in G. When X is E(G), this corresponds to Closed Walk-Vertex Cover (CW-VC, in short). One can similarly define these variants for Feedback Vertex Set, namely Closed Walk-Subgraph Feedback Vertex Set (CW-SFVS, in short) and Closed Walk-Feedback Vertex Set (CW-FVS, in short). Our results are as follows: CW-VC and CW-SVC both admit a polynomial kernel, in contrast to Connected Vertex Cover that does not admit a polynomial kernel unless NP⊆ coNP/ poly.CW-FVS admits a polynomial kernel. On the other hand CW-SFVS does not admit even a polynomial Turing kernel unless the polynomial-time hierarchy collapses. We complement our kernelization algorithms by designing single-exponential FPT algorithms – 2 O ( k )nO ( 1 ) – for all the problems mentioned above. © 2021, Springer Nature Switzerland AG.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

Circumventing Connectivity for Kernelization

We study the problem of approximating a center under the Ulam metric. The Ulam metric, defined over a set of permutations over [n], is the minimum number of move operations (deletion plus insertion) to transform one permutation into another. The Ulam metric is a simpler variant of the general edit distance metric. It provides a measure of dissimilarity over a set of rankings/permutations. In the center problem, given a set of permutations, we are asked to find a permutation (not necessarily from the input set) that minimizes the maximum distance to the input permutations. This problem is also referred to as maximum rank aggregation under Ulam. So far, we only know of a folklore 2-approximation algorithm for this NP-hard problem. Even for constantly many permutations, we do not know anything better than an exhaustive search over all n! permutations. In this paper, we achieve a (32 - 31m)-approximation of the Ulam center in time nO(m2 ln m), for m input permutations over [n]. We therefore get a polynomial time bound while achieving better than a 3/2-approximation for constantly many permutations. This problem is of special interest even for constantly many permutations because under certain dissimilarity measures over rankings, even for four permutations, the problem is NP-hard. In proving our result, we establish a surprising connection between the approximate Ulam center problem and the closest string with wildcards problem (the center problem over the Hamming metric, allowing wildcards). We further study the closest string with wildcards problem and show that there cannot exist any (2 - ϵ)-approximation algorithm (for any ϵ &gt; 0) for it unless P = NP. This inapproximability result is in sharp contrast with the same problem without wildcards, where we know of a PTAS. © Diptarka Chakraborty, Kshitij Gajjar, and Agastya Vibhuti Jha.