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

Pose of an UAV has been traditionally estimated by On Board Computers (OBCs) using Inertial Measurement Unit (IMU) sensor data as input. In this paper, the development of an architecture for estimating the pose of an UAV using popular Computer Vision based methods has bis proposed. Which consists of three sub modules, namely, Image Segmentation Block (IS), Perspective Transform (PT) and Pose Determination (PD) respectively. IS block uses segmentation to detect salient points from the image where as the PT block transforms image coordinates to world coordinates, finally the PD block uses camera parameters and object dimensions to provide Attitude and Translation matrix of the UAV. The proposed approach can adjust with change in environmental parameters. The system was characterized by observing the error in the estimated yaw and estimated depth. Analysis was made on the nature and variation of error with various experimental parameters. An in depth analysis of the paper was carried out. We were finally able to devise an algorithm for estimating the pose of a body without establishing any communication with the body. © 2020, Springer Nature Singapore Pte Ltd.

World University root node

Communications in Computer and Information Science

Pose Estimation of UAVs Using Stereovision

Depth recovery from an image uses popular methods such as stereo-vision, defocus variation, focus or defocus variation, aperture variation, etc. Most of these methods demands extensive computational efforts. Further, the depth can also be obtained from a single image. Depth from single image requires less computational power and is simple to implement. In this work we propose an efficient method that relies on defocus or blur variation in an image to indicate the depth map for a given camera focus. The depth is derived by applying Gaussian filters on the contrast changes or edges in the image. The actual blur amount at the edges is then derived from the ratio of gradient magnitudes of these filtered images. This blur represents a depth map of the actual image. Our method differs from all other methods available in the literature and is based on the fact that we use real data, simple calibration method to extract the depth maps. This is unlike other authors who have mostly relied upon simulated datasets. A unique target is used to calibrate the derived depths by finding coefficients k between theoretical and actual blur. Additionally, the target also characterizes the blur range. © Springer Nature Singapore Pte Ltd 2020.

Calibration of depth map using a novel target

The synthesized Mn2FeAl alloys crystallize in a geometrically frustrated cubic beta-Mn structure (space group: P4(1)32) with an antiferromagnetic ordering whereas the previous theoretical findings suggest for a Heusler structure (L2(1): regular and X: inverse). The experimental stability of the structure is verified by electronic structure calculations performed for various arrangements of Mn, Fe and Al atoms in the beta-Mn-type crystal structure. When compared the energy of the beta-Mn structure with the energy of L2(1) and X type structures, it is found that for an expansion of the lattice volume beta-Mn structure becomes more preferable in total energy than L2(1) and X-type structures. The calculated theoretical equilibrium lattice parameter value for the beta-Mn-2 FeAl is within the accuracy of the experimental value obtained in this work. Additional DFT + U calculations for the optimized crystal structure of the beta-Mn2FeAl revealed that the electronic correlations in the Mn ions result in the increased total magnetic moment. In the X type structure, Mn2FeAl is a half metal, whereas the disordered arrangement of atoms in the beta-Mn structure leads to the closure of the semiconductor gap. The Mn2FeAl alloys exhibit antiferromagnetic ordering (T-N approximate to 42 K), which is in excellent agreement with our electronic structure calculations. The detailed analysis of the magnetic and heat capacity measurements suggests a short-range magnetic ordering in the Mn2FeAl alloys. Owing to the strong antiferromagnetic spin fluctuation caused by the geometric frustration in beta-Mn, a large enhancement in the electronic heat capacity is noticed. Mn2FeAl shows the characteristic features of spin glass as verified from the frequency dependent AC susceptibility analysis using critical power law and Vogel-Fulcher law. To the best of our knowledge, this is the first ever report on the theoretically predicted lowest ground state configuration for Mn2FeAl with a beta-Mn structure and the experimental realization of spin glass features in this geometrically frustrated antiferromagnet.

JOURNAL OF MAGNETISM AND MAGNETIC MATERIALS

Structural stability and magnetic properties of Mn2FeAl alloy with a beta-Mn structure

Due to the issues associated with rare-earth elements, there arises a strong need for magnets with properties between those of ferrites and rare-earth magnets that could substitute the latter in selected applications. Here, we produce a high remanent magnetization composite bonded magnet by mixing FeCo nanowire powders with hexaferrite particles. In the first step, metallic nanowires with diameters between 30 and 100 nm and length of at least 2 mu m are fabricated by electrodeposition. The oriented as-synthesized nanowires show remanence ratios above 0.76 and coercivities above 199 kA/m and resist core oxidation up to 300 degrees C due to the existence of a >8 nm thin oxide passivating shell. In the second step, a composite powder is fabricated by mixing the nanowires with hexaferrite particles. After the optimal nanowire diameter and composite composition are selected, a bonded magnet is produced. The resulting magnet presents a 20\% increase in remanence and an enhancement of the energy product of 48\% with respect to a pure hexaferrite (strontium ferrite) magnet. These results put nanowire-ferrite composites at the forefront as candidate materials for alternative magnets for substitution of rare earths in applications that operate with moderate magnet performance.

Fulltext

ACS APPLIED NANO MATERIALS

FeCo Nanowire-Strontium Ferrite Powder Composites for Permanent Magnets with High-Energy Products

The Convolutional Neural Network (CNN) has witnessed success and vast opportunities in the field of deep learning based video compression. Many deep learning models have either outperformed or performed on par with state-of-the-art compression standards like H.264 and HEVC. In this paper, we propose H.264 Inter-frame prediction based video compression approach using Temporal 3-D CNN based encoder and Y-style CNN based decoder. The proposed architecture includes three stages, Temporal 3-D CNN encoder for forward Predicted (P) frame computation, H.264 like Integer Discrete Cosine Transform and scalar quantization for entropy coding and Y-style CNN for P-frame decoding. The experiments are conducted with different training loss functions and different datasets. The results show that the proposed model outperforms the state-of-the-art compression standards with low computational complexity. © 2020 IEEE.

2020 11th International Conference on Computing, Communication and Networking Technologies, ICCCNT 2020

T3D-Y Codec: A Video Compression Framework using Temporal 3-D CNN Encoder and Y-Style CNN Decoder

Integrated circuits may be vulnerable to hardware Trojan attacks during its design or fabrication phases. This article is a case study of the design of a Viterbi decoder and the effect of hardware Trojans on a coded communication system employing the Viterbi decoder. Design of a Viterbi decoder and possible hardware Trojan models for the same are proposed. An FPGA-based implementation of the decoder and the associated Trojan circuits have been discussed. The noise-added encoded input data stream is stored in the block RAM of the FPGA and the decoded data stream is monitored on the PC through an universal asynchronous receiver transmitter interface. The implementation results show that there is barely any change in the LUTs used (0.5%) and power dissipation (3%) due to the insertion of the proposed Trojan circuits, thus establishing the surreptitious nature of the Trojan. In spite of the fact that the Trojans cause negligible changes in the circuit parameters, there are significant changes in the bit error rate (BER) due to the presence of Trojans. In the absence of Trojans, BER drops down to zero for signal to noise rations (SNRs) higher than 6 dB, but with the presence of Trojans, BER doesn't reduce to zero even at a very high SNRs. This is true even with the Trojan being activated only once during the entire duration of the transmission. © 2020 Kakkara et al.

PeerJ Computer Science

A viterbi decoder and its hardware trojan models: An FPGA-based implementation study

Generating photorealistic images of human subjects in any unseen pose have crucial applications in generating a complete appearance model of the subject. However, from a computer vision perspective, this task becomes significantly challenging due to the inability of modelling the data distribution conditioned on pose. Existing works use a complicated pose transformation model with various additional features such as foreground segmentation, human body parsing etc. to achieve robustness that leads to computational overhead. In this work, we propose a simple yet effective pose transformation GAN by utilizing the Residual Learning method without any additional feature learning to generate a given human image in any arbitrary pose. Using effective data augmentation techniques and cleverly tuning the model, we achieve robustness in terms of illumination, occlusion, distortion and scale. We present a detailed study, both qualitative and quantitative, to demonstrate the superiority of our model over the existing methods on two large datasets. © 2020, Springer Nature Singapore Pte Ltd.

A Robust Pose Transformational GAN for Pose Guided Person Image Synthesis

Recent studies have highlighted a major caveat in various high performing automated systems for tasks such as facial analysis (e.g. gender prediction), object classification, and image to caption generation. Several of the existing systems have been shown to yield biased results towards or against a particular subgroup. The biased behavior exhibited by these models when deployed and used in a real world scenario presents with the challenge of automated systems being unfair. In this research, we propose a novel technique, diversity block, for de-biasing existing models without re-training them. The proposed technique requires small amount of training data and can be incorporated with an existing model for addressing the challenge of biased predictions. This is done by adding a diversity block and computing the prediction based on the scores of the original model and the diversity block in order to get a more confident and de-biased prediction. The efficacy of the proposed technique has been demonstrated on the task of gender prediction, along with an auxiliary case study on object classification. © 2020 IEEE.