arXiv subjects

J. Krishna Murthy

Publications and source records attributed to J. Krishna Murthy.

13 recordsLinked to original sources

arXiv2019

CalibNet: Geometrically Supervised Extrinsic Calibration using 3D Spatial Transformer Networks

3D LiDARs and 2D cameras are increasingly being used alongside each other in sensor rigs for perception tasks. Before these sensors can be used to gather meaningful data, however, their extrinsics (and intrinsics) need to be accurately calibrated, as the performance of the sensor rig is extremely sensitive to these calibration parameters. A vast majority of existing calibration techniques require significant amounts of data and/or calibration targets and human effort, severely impacting their applicability in large-scale production systems. We address this gap with CalibNet: a self-supervised deep network capable of automatically estimating the 6-DoF rigid body transformation between a 3D LiDAR and a 2D camera in real-time. CalibNet alleviates the need for calibration targets, thereby resulting in significant savings in calibration efforts. During training, the network only takes as input a LiDAR point cloud, the corresponding monocular image, and the camera calibration matrix K. At train time, we do not impose direct supervision (i.e., we do not directly regress to the calibration parameters, for example). Instead, we train the network to predict calibration parameters that maximize the geometric and photometric consistency of the input images and point clouds. CalibNet learns to iteratively solve the underlying geometric problem and accurately predicts extrinsic calibration parameters for a wide range of mis-calibrations, without requiring retraining or domain adaptation. The project page is hosted at https://epiception.github.io/CalibNet

cs.RO↗

arXiv2018

Beyond Pixels: Leveraging Geometry and Shape Cues for Online Multi-Object Tracking

This paper introduces geometry and object shape and pose costs for multi-object tracking in urban driving scenarios. Using images from a monocular camera alone, we devise pairwise costs for object tracks, based on several 3D cues such as object pose, shape, and motion. The proposed costs are agnostic to the data association method and can be incorporated into any optimization framework to output the pairwise data associations. These costs are easy to implement, can be computed in real-time, and complement each other to account for possible errors in a tracking-by-detection framework. We perform an extensive analysis of the designed costs and empirically demonstrate consistent improvement over the state-of-the-art under varying conditions that employ a range of object detectors, exhibit a variety in camera and object motions, and, more importantly, are not reliant on the choice of the association framework. We also show that, by using the simplest of associations frameworks (two-frame Hungarian assignment), we surpass the state-of-the-art in multi-object-tracking on road scenes. More qualitative and quantitative results can be found at the following URL: https://junaidcs032.github.io/Geometry_ObjectShape_MOT/.

cs.RO↗

arXiv2018

Geometric Consistency for Self-Supervised End-to-End Visual Odometry

With the success of deep learning based approaches in tackling challenging problems in computer vision, a wide range of deep architectures have recently been proposed for the task of visual odometry (VO) estimation. Most of these proposed solutions rely on supervision, which requires the acquisition of precise ground-truth camera pose information, collected using expensive motion capture systems or high-precision IMU/GPS sensor rigs. In this work, we propose an unsupervised paradigm for deep visual odometry learning. We show that using a noisy teacher, which could be a standard VO pipeline, and by designing a loss term that enforces geometric consistency of the trajectory, we can train accurate deep models for VO that do not require ground-truth labels. We leverage geometry as a self-supervisory signal and propose "Composite Transformation Constraints (CTCs)", that automatically generate supervisory signals for training and enforce geometric consistency in the VO estimate. We also present a method of characterizing the uncertainty in VO estimates thus obtained. To evaluate our VO pipeline, we present exhaustive ablation studies that demonstrate the efficacy of end-to-end, self-supervised methodologies to train deep models for monocular VO. We show that leveraging concepts from geometry and incorporating them into the training of a recurrent neural network results in performance competitive to supervised deep VO methods.

cs.RO↗

arXiv2018

The Earth ain't Flat: Monocular Reconstruction of Vehicles on Steep and Graded Roads from a Moving Camera

Accurate localization of other traffic participants is a vital task in autonomous driving systems. State-of-the-art systems employ a combination of sensing modalities such as RGB cameras and LiDARs for localizing traffic participants, but most such demonstrations have been confined to plain roads. We demonstrate, to the best of our knowledge, the first results for monocular object localization and shape estimation on surfaces that do not share the same plane with the moving monocular camera. We approximate road surfaces by local planar patches and use semantic cues from vehicles in the scene to initialize a local bundle-adjustment like procedure that simultaneously estimates the pose and shape of the vehicles, and the orientation of the local ground plane on which the vehicle stands as well. We evaluate the proposed approach on the KITTI and SYNTHIA-SF benchmarks, for a variety of road plane configurations. The proposed approach significantly improves the state-of-the-art for monocular object localization on arbitrarily-shaped roads.

cs.RO↗

arXiv2018

Constructing Category-Specific Models for Monocular Object-SLAM

We present a new paradigm for real-time object-oriented SLAM with a monocular camera. Contrary to previous approaches, that rely on object-level models, we construct category-level models from CAD collections which are now widely available. To alleviate the need for huge amounts of labeled data, we develop a rendering pipeline that enables synthesis of large datasets from a limited amount of manually labeled data. Using data thus synthesized, we learn category-level models for object deformations in 3D, as well as discriminative object features in 2D. These category models are instance-independent and aid in the design of object landmark observations that can be incorporated into a generic monocular SLAM framework. Where typical object-SLAM approaches usually solve only for object and camera poses, we also estimate object shape on-the-fly, allowing for a wide range of objects from the category to be present in the scene. Moreover, since our 2D object features are learned discriminatively, the proposed object-SLAM system succeeds in several scenarios where sparse feature-based monocular SLAM fails due to insufficient features or parallax. Also, the proposed category-models help in object instance retrieval, useful for Augmented Reality (AR) applications. We evaluate the proposed framework on multiple challenging real-world scenes and show --- to the best of our knowledge --- first results of an instance-independent monocular object-SLAM system and the benefits it enjoys over feature-based SLAM methods.

cs.RO↗

arXiv2016

Reconstructing Vechicles from a Single Image: Shape Priors for Road Scene Understanding

We present an approach for reconstructing vehicles from a single (RGB) image, in the context of autonomous driving. Though the problem appears to be ill-posed, we demonstrate that prior knowledge about how 3D shapes of vehicles project to an image can be used to reason about the reverse process, i.e., how shapes (back-)project from 2D to 3D. We encode this knowledge in \emph{shape priors}, which are learnt over a small keypoint-annotated dataset. We then formulate a shape-aware adjustment problem that uses the learnt shape priors to recover the 3D pose and shape of a query object from an image. For shape representation and inference, we leverage recent successes of Convolutional Neural Networks (CNNs) for the task of object and keypoint localization, and train a novel cascaded fully-convolutional architecture to localize vehicle \emph{keypoints} in images. The shape-aware adjustment then robustly recovers shape (3D locations of the detected keypoints) while simultaneously filling in occluded keypoints. To tackle estimation errors incurred due to erroneously detected keypoints, we use an Iteratively Re-weighted Least Squares (IRLS) scheme for robust optimization, and as a by-product characterize noise models for each predicted keypoint. We evaluate our approach on autonomous driving benchmarks, and present superior results to existing monocular, as well as stereo approaches.

cs.CV↗

arXiv2016

Absence of dipolar ordering in Co doped CuO

Polycrystalline CuO samples with Co doping were prepared by solid state method with flowing oxygen condition and examined their structural and multiferroic properties. Structural studies have confirmed single phase monoclinic crystal structure of all samples, however, in Co doped samples a decrease in volume with an increase in monoclinic distortion is found. For pristine sample, temperature dependent magnetization has confirmed two antiferromagnetic (AFM) transitions at 213 K and 230 K and frequency independent dielectric peaks at these AFM transitions suggesting the ferroelectric nature. Magnetization of the Co doped samples has showed a marginal increase in ordering temperature of the high-temperature AFM transition and decrease in low temperature AFM ordering temperature. Further, doped samples have shown giant dielectric constant with no signature of ferroelectricity. The x-ray photoelectric spectroscopy study has revealed multiple valance states for both Co and Cu in the doped samples that simultaneously explain the giant dielectric constant and suppression of ferroelectric order.

cond-mat.str-el↗

arXiv2015

Giant magnetocaloric effect in Gd2NiMnO6 and Gd2CoMnO6 ferromagnetic insulators

We have investigated magnetocaloric effect in double perovskite Gd2NiMnO6 (GNMO) and Gd2CoMnO6 (GCMO) samples by magnetic and heat capacity measurements. Ferromagnetic ordering is observed at ~130 K (~112 K) in GNMO (GCMO), while the Gd exchange interactions seem to dominate for T < 20 K. In GCMO, below 50 K, an antiferromagnetic behaviour due to 3d-4f exchnage interaction is observed. A maximum entropy (-ΔSM) and adiabatic temperature change of ~35.5 J Kg-1 K-1 (~24 J Kg-1 K-1) and 10.5 K (6.5 K) is observed in GNMO (GCMO) for a magnetic field change of 7 T at low temperatures. Absence of magnetic and thermal hysteresis and their insulating nature make them promising for low temperature magnetic refrigeration.

cond-mat.mtrl-sci↗

arXiv2015

Antisite disorder driven spontaneous exchange bias effect in La2-xSrxCoMnO6 (0<x<1)

Doping at the rare-earth site by divalent alkaline-earth ions in perovskite lattice has witnessed a variety of magnetic and electronic orders with spatially correlated charge, spin and orbital degrees of freedom. Here, we report an antisite disorder driven spontaneous exchange bias effect as a result of hole carrier (Sr2+) doping in La2-xSrxCoMnO6 (0 < x < 1) double perovskites. X-ray diffraction and Raman spectroscopy have evidenced an increase in disorder with the increase of Sr content up to x = 0.5 and thereby decreases from x = 0.5 to 1. X-ray absorption spectroscopy has revealed that only Co is present in mixed valent Co2+ and Co3+ states with Sr doping to compensate the charge neutrality. Magnetotransport is strongly correlated with the increase of antisite disorder. The antisite disorder at the B-site interrupts the long-range ferromagnetic order by introducing various magnetic interactions and instigates reentrant glassy dynamics, phase separation and canted type antiferromagnetic behavior with the decrease of temperature. This leads to novel magnetic microstructure with unidirectional anisotropy that causes spontaneous exchange bias effect that can be tuned with the amount of antisite disorder.

cond-mat.str-el↗

arXiv2014

Metamagnetic behavior and effect of field cooling on sharp magnetization jumps in multiferroic Y2CoMnO6

We present sharp magnetization jumps and field induced irreversibility in magnetization in multiferroic Y2CoMnO6. Appearance of magnetic relaxation and field sweep rate dependence of magnetization jumps resemble the martensite like scenario and suggests the coexistence of E*-type antiferromagnetic and ferromagnetic phases at low temperatures. In Y2CoMnO6, the critical field required for the sharp jump can be increased or decreased depening on the magnitude and direction of the cooling field; this is remarkably different from manganites or other metamagnetic materials where the critical field increases irrespective of the direction of the field cooling. The cooling field dependence on the sharp magnetization jumps has been described by considering exchange pinning mechanism at the interface, like in exchange bias model.

cond-mat.str-el↗

arXiv2014

Multicaloric effect in multiferroic Y2CoMnO6

We have investigated multiple caloric effects in multiferroic Y2CoMnO6. Polycrystalline sample prepared by solid state method has shown a ferromagnetic Curie temperature 75 K with second order phase transition; a maximum magneto entropy change -$Δ$S_{Mmax}) of ~ 7.3 J/kg K with reasonable relative cooling power 220 J/kg is found without thermal and magnetic hysteresis loss. Electric field driven entropy change (-$Δ$S_{Emax}) of ~ 0.26 J/m^3 K obtained using the Maxwells relation and estimated magnetically induced total temperature change of 5.45 K around Curie temperature that confirms the multicaloric effect in the sample.

cond-mat.mtrl-sci↗

arXiv2014

Giant zero field cooled spontaneous exchange bias effect in phase separated La1.5Sr0.5CoMnO6

We report a giant zero field cooled exchange bias (ZEB) effect (~0.65 T) in La1.5Sr0.5CoMnO6 sample. Magnetic study has revealed a reentrant spin glass ~90 K, phase separation to spin glass and ferromagnetic phases below 50 K and canted antiferromagnetic transition ~10 K. A small conventional exchange bias (CEB) is established with the advent of spontaneous phase separation down to 10 K. Giant ZEB and enhanced CEB effects are found only below 10 K and are attributed to the large unidirectional anisotropy at the interface of isothermally field induced ferromagnetic phase and canted antiferromagnetic background.

cond-mat.mtrl-sci↗

arXiv2012

Magnetodielectric behavior in La2CoMnO6 nanoparticles

We have investigated magnetic, dielectric and magnetodielectric properties of La2CoMnO6 nanoparticles prepared by sol-gel method. Magnetization measurements revealed two distinct ferromagnetic transitions at 218 K and 135 K that can be assigned to ordered and disordered magnetic phases of the La2CoMnO6 nanoparticles. Two dielectric relaxations culminating around the magnetic transitions were observed with a maximum magnetodielectric response reaching 10% and 8% at the respective relaxation peaks measured at 100 kHz under 5T magnetic field. The dc electrical resistivity followed an insulating behavior and showed a negative magnetoresistance; there was no noticeable anomaly in resistivity or magnetoresistance near the magnetic ordering temperatures. Complex impedance analysis revealed a clear intrinsic contribution to the magnetodielectric response; however, extrinsic contribution due to Maxwell-Wagner effect combined with magnetoresistance property dominated the magnetodielectric effect at high temperatures.

cond-mat.mtrl-sci↗