arXiv subjects

Sudip Das

Publications and source records attributed to Sudip Das.

15 recordsLinked to original sources

arXiv2024

Reflective Teacher: Semi-Supervised Multimodal 3D Object Detection in Bird's-Eye-View via Uncertainty Measure

Applying pseudo labeling techniques has been found to be advantageous in semi-supervised 3D object detection (SSOD) in Bird's-Eye-View (BEV) for autonomous driving, particularly where labeled data is limited. In the literature, Exponential Moving Average (EMA) has been used for adjustments of the weights of teacher network by the student network. However, the same induces catastrophic forgetting in the teacher network. In this work, we address this issue by introducing a novel concept of Reflective Teacher where the student is trained by both labeled and pseudo labeled data while its knowledge is progressively passed to the teacher through a regularizer to ensure retention of previous knowledge. Additionally, we propose Geometry Aware BEV Fusion (GA-BEVFusion) for efficient alignment of multi-modal BEV features, thus reducing the disparity between the modalities - camera and LiDAR. This helps to map the precise geometric information embedded among LiDAR points reliably with the spatial priors for extraction of semantic information from camera images. Our experiments on the nuScenes and Waymo datasets demonstrate: 1) improved performance over state-of-the-art methods in both fully supervised and semi-supervised settings; 2) Reflective Teacher achieves equivalent performance with only 25% and 22% of labeled data for nuScenes and Waymo datasets respectively, in contrast to other fully supervised methods that utilize the full labeled dataset.

cs.CV↗

arXiv2023

UnShadowNet: Illumination Critic Guided Contrastive Learning For Shadow Removal

Shadows are frequently encountered natural phenomena that significantly hinder the performance of computer vision perception systems in practical settings, e.g., autonomous driving. A solution to this would be to eliminate shadow regions from the images before the processing of the perception system. Yet, training such a solution requires pairs of aligned shadowed and non-shadowed images which are difficult to obtain. We introduce a novel weakly supervised shadow removal framework UnShadowNet trained using contrastive learning. It is composed of a DeShadower network responsible for the removal of the extracted shadow under the guidance of an Illumination network which is trained adversarially by the illumination critic and a Refinement network to further remove artefacts. We show that UnShadowNet can be easily extended to a fully-supervised set-up to exploit the ground-truth when available. UnShadowNet outperforms existing state-of-the-art approaches on three publicly available shadow datasets (ISTD, adjusted ISTD, SRD) in both the weakly and fully supervised setups.

cs.CV↗

arXiv2023

Revisiting Modality Imbalance In Multimodal Pedestrian Detection

Multimodal learning, particularly for pedestrian detection, has recently received emphasis due to its capability to function equally well in several critical autonomous driving scenarios such as low-light, night-time, and adverse weather conditions. However, in most cases, the training distribution largely emphasizes the contribution of one specific input that makes the network biased towards one modality. Hence, the generalization of such models becomes a significant problem where the non-dominant input modality during training could be contributing more to the course of inference. Here, we introduce a novel training setup with regularizer in the multimodal architecture to resolve the problem of this disparity between the modalities. Specifically, our regularizer term helps to make the feature fusion method more robust by considering both the feature extractors equivalently important during the training to extract the multimodal distribution which is referred to as removing the imbalance problem. Furthermore, our decoupling concept of output stream helps the detection task by sharing the spatial sensitive information mutually. Extensive experiments of the proposed method on KAIST and UTokyo datasets shows improvement of the respective state-of-the-art performance.

cs.CV↗

arXiv2022

Deep Multi-Task Networks For Occluded Pedestrian Pose Estimation

Most of the existing works on pedestrian pose estimation do not consider estimating the pose of an occluded pedestrian, as the annotations of the occluded parts are not available in relevant automotive datasets. For example, CityPersons, a well-known dataset for pedestrian detection in automotive scenes does not provide pose annotations, whereas MS-COCO, a non-automotive dataset, contains human pose estimation. In this work, we propose a multi-task framework to extract pedestrian features through detection and instance segmentation tasks performed separately on these two distributions. Thereafter, an encoder learns pose specific features using an unsupervised instance-level domain adaptation method for the pedestrian instances from both distributions. The proposed framework has improved state-of-the-art performances of pose estimation, pedestrian detection, and instance segmentation.

cs.CV↗

arXiv2022

Improving self-supervised pretraining models for epileptic seizure detection from EEG data

There is abundant medical data on the internet, most of which are unlabeled. Traditional supervised learning algorithms are often limited by the amount of labeled data, especially in the medical domain, where labeling is costly in terms of human processing and specialized experts needed to label them. They are also prone to human error and biased as a select few expert annotators label them. These issues are mitigated by Self-supervision, where we generate pseudo-labels from unlabelled data by seeing the data itself. This paper presents various self-supervision strategies to enhance the performance of a time-series based Diffusion convolution recurrent neural network (DCRNN) model. The learned weights in the self-supervision pretraining phase can be transferred to the supervised training phase to boost the model's prediction capability. Our techniques are tested on an extension of a Diffusion Convolutional Recurrent Neural network (DCRNN) model, an RNN with graph diffusion convolutions, which models the spatiotemporal dependencies present in EEG signals. When the learned weights from the pretraining stage are transferred to a DCRNN model to determine whether an EEG time window has a characteristic seizure signal associated with it, our method yields an AUROC score $1.56\%$ than the current state-of-the-art models on the TUH EEG seizure corpus.

eess.SP↗

arXiv2022

Spatio-Contextual Deep Network Based Multimodal Pedestrian Detection For Autonomous Driving

Pedestrian Detection is the most critical module of an Autonomous Driving system. Although a camera is commonly used for this purpose, its quality degrades severely in low-light night time driving scenarios. On the other hand, the quality of a thermal camera image remains unaffected in similar conditions. This paper proposes an end-to-end multimodal fusion model for pedestrian detection using RGB and thermal images. Its novel spatio-contextual deep network architecture is capable of exploiting the multimodal input efficiently. It consists of two distinct deformable ResNeXt-50 encoders for feature extraction from the two modalities. Fusion of these two encoded features takes place inside a multimodal feature embedding module (MuFEm) consisting of several groups of a pair of Graph Attention Network and a feature fusion unit. The output of the last feature fusion unit of MuFEm is subsequently passed to two CRFs for their spatial refinement. Further enhancement of the features is achieved by applying channel-wise attention and extraction of contextual information with the help of four RNNs traversing in four different directions. Finally, these feature maps are used by a single-stage decoder to generate the bounding box of each pedestrian and the score map. We have performed extensive experiments of the proposed framework on three publicly available multimodal pedestrian detection benchmark datasets, namely KAIST, CVC-14, and UTokyo. The results on each of them improved the respective state-of-the-art performance. A short video giving an overview of this work along with its qualitative results can be seen at https://youtu.be/FDJdSifuuCs. Our source code will be released upon publication of the paper.

cs.CV↗

arXiv2021

Effect of pulse width on the dynamics of a deflated vesicle in unipolar and bipolar pulsed electric fields

Giant unilamellar vesicles subjected to pulsed direct-current (pulsed-DC) fields are promising biomimetic systems to investigate the electroporation of cells. In strong electric fields, vesicles undergo significant deformation, which strongly alters the transmembrane potential, consequently the electroporation. Previous theoretical studies investigated the electrodeformation of vesicles in DC fields (which are not pulsed). In this work, we computationally investigate the deformation of a deflated vesicle under unipolar, bipolar, and two-step unipolar pulses and show sensitive dependence of intermediate shapes on type of pulse and the pulse width. Starting with the stress-free initial shape of a deflated vesicle, which is similar to a prolate spheroid, the analysis is presented for the cases with higher and lower conductivities of the inner fluid medium relative to the outer fluid medium. For the ratio of inner to outer fluid conductivity, $σ_\mathrm{r}$ = 10, the shape always remains prolate, including when the field is turned off. For $σ_\mathrm{r} = 0.1$, several complex dynamics are observed, such as the prolate-to-oblate (PO), prolate-to-oblate-to-prolate (POP) shape transitions in time depending upon the strength of the field and the pulse properties. In this case, on turning off the field, a metastable oblate equilibrium shape is seen, that seems to be a characteristics of a deflated vesicle leading to POPO transitions. When a two-step unipolar pulse (a combination of a strong and a weak subpulse) is applied, a vesicle can reach an oblate or a prolate final shape depending upon the relative durations of the two subpulses.

cond-mat.soft↗

arXiv2020

Shock related unsteadiness of axisymmetric spiked bodies in the supersonic flow

Shock related unsteadiness over axisymmetric spiked body configurations is experimentally investigated at a freestream supersonic Mach number of 2.0 at 0$^\circ$ angle of attack. Three different forebody configurations mounted with a sharp spike-tip ranging from blunt to streamlined (flat-face, hemispherical, and elliptical) are considered. Steady and unsteady pressure measurements, short-exposure high-speed shadowgraphy, shock footprint analysis from $x-t$ plots, and identification of dominant spatiotemporal modes through modal analysis are carried out to explain the unsteady flow physics. The present investigation tools are validated against the well-known events of `pulsation' associated with the flat-face case. The hemispherical case is characterized by the formation of a separated free shear layer and associated localized shock oscillations. The cycle of charging and ejection of fluid mass from the recirculation zone, confined between the separated shear layer and the spiked body, is identified to drive the flow unsteadiness. Such an event triggers the out-of-phase motion between the separated and reattachment shocks. In the elliptical case, the overall flow field resembles that of the hemispherical case, except with dampened unsteadiness. The value of the cone angle ($λ$) associated with the recirculation region is found to be responsible for the fluctuations from the charging and ejection of fluid mass. Thereby it controls the extent of out-of-shock phase motion. In the elliptical case, $λ$ is observed to be smaller and exhibits a reduction in shock unsteadiness. \hlt{Based on the gathered results and understanding, the reduction in unsteadiness associated with the aerodisk mounted on the hemispherical forebody is explained via the almost complete elimination of the out-of-phase shock motion.

physics.flu-dyn↗

arXiv2020

Parametric experimental studies on the shock related unsteadiness in a hemispherical spiked body at supersonic flow

Experimental studies are carried out to investigate the effects of the geometrical parameters with a drag reducing spike on a hemispherical forebody in a supersonic freestream of $M_\infty=2.0$ at $0^{\circ}$ angle of attack. The spike length $(l/D=0.5,1.0,1.5,2.0)$, spike stem diameter $(d/D=0.06,0.12,0.18)$, and spike tip shapes are varied and their influence on the time-averaged, and time-resolved flow field are examined. When $l/D$ increases, a significant reduction in drag ($c_d$) is achieved at $l/D=1.5$, whereas the variation in $d/D$ has only a minor effect. The intensity of the shock-related unsteadiness is reduced with an increase in {$d/D$ from $0.06$ to $0.18$}, whereas changes in $l/D$ have a negligible effect. The effects of spike tip geometry are studied by replacing the sharp spike tip with a hemispherical one having three different base shapes (vertical base, circular base, and elliptical base). Hemispherical spike tip with a vertical base is performing better by reducing $c_d$ and flow unsteadiness. The dominant Spatio-temporal mode arising due to the shock-related unsteadiness is represented through modal analysis of time-resolved shadowgraph images and the findings are consistent with the other measurements.

physics.flu-dyn↗

arXiv2020

Scale-Invariant Multi-Oriented Text Detection in Wild Scene Images

Automatic detection of scene texts in the wild is a challenging problem, particularly due to the difficulties in handling (i) occlusions of varying percentages, (ii) widely different scales and orientations, (iii) severe degradations in the image quality etc. In this article, we propose a fully convolutional neural network architecture consisting of a novel Feature Representation Block (FRB) capable of efficient abstraction of information. The proposed network has been trained using curriculum learning with respect to difficulties in image samples and gradual pixel-wise blurring. It is capable of detecting texts of different scales and orientations suffered by blurring from multiple possible sources, non-uniform illumination as well as partial occlusions of varying percentages. Text detection performance of the proposed framework on various benchmark sample databases including ICDAR 2015, ICDAR 2017 MLT, COCO-Text and MSRA-TD500 improves respective state-of-the-art results significantly. Source code of the proposed architecture will be made available at github.

cs.CV↗

arXiv2020

An End-to-End Framework for Unsupervised Pose Estimation of Occluded Pedestrians

Pose estimation in the wild is a challenging problem, particularly in situations of (i) occlusions of varying degrees and (ii) crowded outdoor scenes. Most of the existing studies of pose estimation did not report the performance in similar situations. Moreover, pose annotations for occluded parts of human figures have not been provided in any of the relevant standard datasets which in turn creates further difficulties to the required studies for pose estimation of the entire figure of occluded humans. Well known pedestrian detection datasets such as CityPersons contains samples of outdoor scenes but it does not include pose annotations. Here, we propose a novel multi-task framework for end-to-end training towards the entire pose estimation of pedestrians including in situations of any kind of occlusion. To tackle this problem for training the network, we make use of a pose estimation dataset, MS-COCO, and employ unsupervised adversarial instance-level domain adaptation for estimating the entire pose of occluded pedestrians. The experimental studies show that the proposed framework outperforms the SOTA results for pose estimation, instance segmentation and pedestrian detection in cases of heavy occlusions (HO) and reasonable + heavy occlusions (R + HO) on the two benchmark datasets.

cs.CV↗

arXiv2019

Seek and You Will Find: A New Optimized Framework for Efficient Detection of Pedestrian

Studies of object detection and localization, particularly pedestrian detection have received considerable attention in recent times due to its several prospective applications such as surveillance, driving assistance, autonomous cars, etc. Also, a significant trend of latest research studies in related problem areas is the use of sophisticated Deep Learning based approaches to improve the benchmark performance on various standard datasets. A trade-off between the speed (number of video frames processed per second) and detection accuracy has often been reported in the existing literature. In this article, we present a new but simple deep learning based strategy for pedestrian detection that improves this trade-off. Since training of similar models using publicly available sample datasets failed to improve the detection performance to some significant extent, particularly for the instances of pedestrians of smaller sizes, we have developed a new sample dataset consisting of more than 80K annotated pedestrian figures in videos recorded under varying traffic conditions. Performance of the proposed model on the test samples of the new dataset and two other existing datasets, namely Caltech Pedestrian Dataset (CPD) and CityPerson Dataset (CD) have been obtained. Our proposed system shows nearly 16\% improvement over the existing state-of-the-art result.

cs.CV↗

arXiv2018

Deformation of a biconcave-discoid capsule in extensional flow and electric field

Biconcave-discoid (empirical shape of a red blood cell) capsule finds numerous applications in the field of bio-fluid dynamics and rheology, apart from understanding the behavior of red blood cells (RBC) in blood flow. A detailed analysis is, therefore, carried out to understand the effects of the uniaxial extensional flow and electric field on the deformation of a RBC and biconcave-discoid polymeric capsule in the axisymmetric regime. The transient deformation is computed numerically using axisymmetric boundary integral method for Stokes flow considering the Skalak membrane constitutive law as the model for the area incompressible RBC/biconcave-discoid capsule membrane. A remarkable biconcave-discoid to prolate spheroid transition is observed when the elastic energy is overcome by the viscous or Maxwell electric stresses. Moreover, the significance of membrane stresses developed during the deformation and at steady state and different modes of deformation are presented. This study should be useful in designing an artificial system involving biological cells under an electric field.

cond-mat.soft↗

arXiv2017

Large deformation electrohydrodynamics of a Skalak elastic capsule in AC electric field

The axisymmetric electrohydrodynamic deformation of an elastic capsule with capacitive membrane obeying Skalak law under uniform AC electric field is investigated using analytical and boundary integral theory. The low capillary number (the ratio of destabilizing shear or electric force to the stabilizing elastic force) regime shows time-averaged prolate and oblate spheroid deformations, and the time-periodic prolate-sphere, oblate-sphere breathing modes are commensurate with the time averaged-deformation. A novel prolate-oblate breathing mode is observed due to an interplay of finite membrane charging time and the field reversal of the AC field. The study, when extended to high capillary number, shows new breathing modes of cylinder-prolate, cylinder-oblate, and biconcave-prolate deformation. These are the results of highly compressive normal Maxwell stress at the poles and are aided by weak compressive equatorial stress, characteristic of a capacitive membrane. The findings of this work should form the basis for the understanding of more complex biological cells and synthetic capsules for industrial applications.

cond-mat.soft↗

arXiv2017

Large deformation electrohydrodynamics of an elastic capsule in DC electric field

The dynamics of a spherical elastic capsule, containing a Newtonian fluid bounded by an elastic membrane and immersed in another Newtonian fluid, in a uniform DC electric field is investigated. Discontinuity of electrical properties such as conductivities of the internal and external fluid media as well as capacitance and conductance of the membrane lead to a net interfacial Maxwell stress which can cause the deformation of such an elastic capsule. We investigate this problem considering well established membrane laws for a thin elastic membrane, with fully resolved hydrodynamics in the Stokes flow limit and describe the electrostatics using the capacitor model. In the limit of small deformation, the analytical theory predicts the dynamics fairly satisfactorily. Large deformations at high capillary number though necessitate a numerical approach (Boundary element method in the present case) to solve this highly non-linear problem. Akin to vesicles, at intermediate times, highly nonlinear biconcave shapes along with squaring and hexagon like shapes are observed when the outer medium is more conducting. The study identifies the essentiality of parameters such as high membrane capacitance, low membrane conductance, low hydrodynamic time scales and high capillary number for observation of these shape transitions. The transition is due to large compressive Maxwell stress at the poles at intermediate times. Thus such shape transition can be seen in spherical globules admitting electrical capacitance, possibly, irrespective of the nature of the interfacial restoring force.

cond-mat.soft↗