arXiv subjects

Brejesh Lall

Publications and source records attributed to Brejesh Lall.

71 records · Page 4Linked to original sources

arXiv2019

DSAL-GAN: Denoising based Saliency Prediction with Generative Adversarial Networks

Synthesizing high quality saliency maps from noisy images is a challenging problem in computer vision and has many practical applications. Samples generated by existing techniques for saliency detection cannot handle the noise perturbations smoothly and fail to delineate the salient objects present in the given scene. In this paper, we present a novel end-to-end coupled Denoising based Saliency Prediction with Generative Adversarial Network (DSAL-GAN) framework to address the problem of salient object detection in noisy images. DSAL-GAN consists of two generative adversarial-networks (GAN) trained end-to-end to perform denoising and saliency prediction altogether in a holistic manner. The first GAN consists of a generator which denoises the noisy input image, and in the discriminator counterpart we check whether the output is a denoised image or ground truth original image. The second GAN predicts the saliency maps from raw pixels of the input denoised image using a data-driven metric based on saliency prediction method with adversarial loss. Cycle consistency loss is also incorporated to further improve salient region prediction. We demonstrate with comprehensive evaluation that the proposed framework outperforms several baseline saliency models on various performance benchmarks.

cs.CV↗

arXiv2019

DeepPoint3D: Learning Discriminative Local Descriptors using Deep Metric Learning on 3D Point Clouds

Learning local descriptors is an important problem in computer vision. While there are many techniques for learning local patch descriptors for 2D images, recently efforts have been made for learning local descriptors for 3D points. The recent progress towards solving this problem in 3D leverages the strong feature representation capability of image based convolutional neural networks by utilizing RGB-D or multi-view representations. However, in this paper, we propose to learn 3D local descriptors by directly processing unstructured 3D point clouds without needing any intermediate representation. The method constitutes a deep network for learning permutation invariant representation of 3D points. To learn the local descriptors, we use a multi-margin contrastive loss which discriminates between similar and dissimilar points on a surface while also leveraging the extent of dissimilarity among the negative samples at the time of training. With comprehensive evaluation against strong baselines, we show that the proposed method outperforms state-of-the-art methods for matching points in 3D point clouds. Further, we demonstrate the effectiveness of the proposed method on various applications achieving state-of-the-art results.

cs.CV↗

arXiv2019

Fast Hierarchical Depth Map Computation from Stereo

Disparity by Block Matching stereo is usually used in applications with limited computational power in order to get depth estimates. However, the research on simple stereo methods has been lesser than the energy based counterparts which promise a better quality depth map with more potential for future improvements. Semi-global-matching (SGM) methods offer good performance and easy implementation but suffer from the problem of very high memory footprint because it's working on the full disparity space image. On the other hand, Block matching stereo needs much less memory. In this paper, we introduce a novel multi-scale-hierarchical block-matching approach using a pyramidal variant of depth and cost functions which drastically improves the results of standard block matching stereo techniques while preserving the low memory footprint and further reducing the complexity of standard block matching. We tested our new multi block matching scheme on the Middlebury stereo benchmark. For the Middlebury benchmark we get results that are only slightly worse than state of the art SGM implementations.

cs.CV↗

arXiv2018

Nrityantar: Pose oblivious Indian classical dance sequence classification system

In this paper, we attempt to advance the research work done in human action recognition to a rather specialized application namely Indian Classical Dance (ICD) classification. The variation in such dance forms in terms of hand and body postures, facial expressions or emotions and head orientation makes pose estimation an extremely challenging task. To circumvent this problem, we construct a pose-oblivious shape signature which is fed to a sequence learning framework. The pose signature representation is done in two-fold process. First, we represent person-pose in first frame of a dance video using symmetric Spatial Transformer Networks (STN) to extract good person object proposals and CNN-based parallel single person pose estimator (SPPE). Next, the pose basis are converted to pose flows by assigning a similarity score between successive poses followed by non-maximal suppression. Instead of feeding a simple chain of joints in the sequence learner which generally hinders the network performance we constitute a feature vector of the normalized distance vectors, flow, angles between anchor joints which captures the adjacency configuration in the skeletal pattern. Thus, the kinematic relationship amongst the body joints across the frames using pose estimation helps in better establishing the spatio-temporal dependencies. We present an exhaustive empirical evaluation of state-of-the-art deep network based methods for dance classification on ICD dataset.

cs.CV↗

arXiv2018

Application of Vector Sensor for Underwater Acoustic Communications

The use of vector sensors as receivers for Underwater Acoustic Communications systems is gaining popularity. It has become important to obtain performance measures for such communication systems to quantify their efficacy. The fundamental advantage of using a vector sensor as a receiver is that a single sensor is able to provide diversity gains offered by MIMO systems. In a recent work novel framework for evaluating capacity of underwater channel was proposed. The approach is based on modeling the channel as a set of paths along which the signal arrives at the receiver with different Angles of Arrival. In this work, we build on that framework to provide a bound on the achievable capacity of such a system. The analytical bounds have been compared against simulation results for a vector sensor based SIMO underwater communications system. The channel parameters are modeled by analysing the statistics generated with Bellhop simulation tool. This representation of the channel is flexible and allows for characterizing channels at differenet geographical locations and at different time instances. This characterization in terms of channel parameters enables the computing of the performance measure (channel capacity bound) for different geographical locations

eess.SP↗

arXiv2018

Object cosegmentation using deep Siamese network

Object cosegmentation addresses the problem of discovering similar objects from multiple images and segmenting them as foreground simultaneously. In this paper, we propose a novel end-to-end pipeline to segment the similar objects simultaneously from relevant set of images using supervised learning via deep-learning framework. We experiment with multiple set of object proposal generation techniques and perform extensive numerical evaluations by training the Siamese network with generated object proposals. Similar objects proposals for the test images are retrieved using the ANNOY (Approximate Nearest Neighbor) library and deep semantic segmentation is performed on them. Finally, we form a collage from the segmented similar objects based on the relative importance of the objects.

cs.CV↗

arXiv2018

Large Scale Novel Object Discovery in 3D

We present a method for discovering never-seen-before objects in 3D point clouds obtained from sensors like Microsoft Kinect. We generate supervoxels directly from the point cloud data and use them with a Siamese network, built on a recently proposed 3D convolutional neural network architecture. We use known objects to train a non-linear embedding of supervoxels, by optimizing the criteria that supervoxels which fall on the same object should be closer than those which fall on different objects, in the embedding space. We test on unknown objects, which were not seen during training, and perform clustering in the learned embedding space of supervoxels to effectively perform novel object discovery. We validate the method with extensive experiments, quantitatively showing that it can discover numerous unseen objects while being trained on only a few dense 3D models. We also show very good qualitative results of object discovery in point cloud data when the test objects, either specific instances or even categories, were never seen during training.

cs.CV↗

arXiv2017

Object Classification using Ensemble of Local and Deep Features

In this paper we propose an ensemble of local and deep features for object classification. We also compare and contrast effectiveness of feature representation capability of various layers of convolutional neural network. We demonstrate with extensive experiments for object classification that the representation capability of features from deep networks can be complemented with information captured from local features. We also find out that features from various deep convolutional networks encode distinctive characteristic information. We establish that, as opposed to conventional practice, intermediate layers of deep networks can augment the classification capabilities of features obtained from fully connected layers.

cs.CV↗

arXiv2017

Enhanced Characterness for Text Detection in the Wild

Text spotting is an interesting research problem as text may appear at any random place and may occur in various forms. Moreover, ability to detect text opens the horizons for improving many advanced computer vision problems. In this paper, we propose a novel language agnostic text detection method utilizing edge enhanced Maximally Stable Extremal Regions in natural scenes by defining strong characterness measures. We show that a simple combination of characterness cues help in rejecting the non text regions. These regions are further fine-tuned for rejecting the non-textual neighbor regions. Comprehensive evaluation of the proposed scheme shows that it provides comparative to better generalization performance to the traditional methods for this task.

cs.CV↗

arXiv2017

Large-Scale 3D Shape Reconstruction and Segmentation from ShapeNet Core55

We introduce a large-scale 3D shape understanding benchmark using data and annotation from ShapeNet 3D object database. The benchmark consists of two tasks: part-level segmentation of 3D shapes and 3D reconstruction from single view images. Ten teams have participated in the challenge and the best performing teams have outperformed state-of-the-art approaches on both tasks. A few novel deep learning architectures have been proposed on various 3D representations on both tasks. We report the techniques used by each team and the corresponding performances. In addition, we summarize the major discoveries from the reported results and possible trends for the future work in the field.

cs.CV↗

arXiv2017

Drought Stress Classification using 3D Plant Models

Quantification of physiological changes in plants can capture different drought mechanisms and assist in selection of tolerant varieties in a high throughput manner. In this context, an accurate 3D model of plant canopy provides a reliable representation for drought stress characterization in contrast to using 2D images. In this paper, we propose a novel end-to-end pipeline including 3D reconstruction, segmentation and feature extraction, leveraging deep neural networks at various stages, for drought stress study. To overcome the high degree of self-similarities and self-occlusions in plant canopy, prior knowledge of leaf shape based on features from deep siamese network are used to construct an accurate 3D model using structure from motion on wheat plants. The drought stress is characterized with a deep network based feature aggregation. We compare the proposed methodology on several descriptors, and show that the network outperforms conventional methods.

cs.CV↗

arXiv2017

3D Binary Signatures

In this paper, we propose a novel binary descriptor for 3D point clouds. The proposed descriptor termed as 3D Binary Signature (3DBS) is motivated from the matching efficiency of the binary descriptors for 2D images. 3DBS describes keypoints from point clouds with a binary vector resulting in extremely fast matching. The method uses keypoints from standard keypoint detectors. The descriptor is built by constructing a Local Reference Frame and aligning a local surface patch accordingly. The local surface patch constitutes of identifying nearest neighbours based upon an angular constraint among them. The points are ordered with respect to the distance from the keypoints. The normals of the ordered pairs of these keypoints are projected on the axes and the relative magnitude is used to assign a binary digit. The vector thus constituted is used as a signature for representing the keypoints. The matching is done by using hamming distance. We show that 3DBS outperforms state of the art descriptors on various evaluation metrics.

cs.CV↗

arXiv2017

Per-Tone model for Common Mode sensor based alien noise cancellation for Downstream xDSL

For xDSL systems, alien noise cancellation using an additional common mode sensor at the downstream receiver can be thought of as interference cancellation in a Single Input Dual Output (SIDO) system. The coupling between the common mode and differential mode can be modelled as an LTI system with a long impulse response, resulting in high complexity for cancellation. Frequency domain per-tone cancellation offers a low complexity approach to the problem besides having other advantages like faster training, but suffers from loss in cancellation performance due to approximations in the per-tone model. We analyze this loss and show that it is possible to minimize it by a convenient post-training "delay" adjustment. We also show via measurements that the loss of cancellation performance due to the per-tone model is not very large for real scenarios.

cs.IT↗

arXiv2017

SalProp: Salient object proposals via aggregated edge cues

In this paper, we propose a novel object proposal generation scheme by formulating a graph-based salient edge classification framework that utilizes the edge context. In the proposed method, we construct a Bayesian probabilistic edge map to assign a saliency value to the edgelets by exploiting low level edge features. A Conditional Random Field is then learned to effectively combine these features for edge classification with object/non-object label. We propose an objectness score for the generated windows by analyzing the salient edge density inside the bounding box. Extensive experiments on PASCAL VOC 2007 dataset demonstrate that the proposed method gives competitive performance against 10 popular generic object detection techniques while using fewer number of proposals.

cs.CV↗

arXiv2016

Ranking academic institutions on potential paper acceptance in upcoming conferences

The crux of the problem in KDD Cup 2016 involves developing data mining techniques to rank research institutions based on publications. Rank importance of research institutions are derived from predictions on the number of full research papers that would potentially get accepted in upcoming top-tier conferences, utilizing public information on the web. This paper describes our solution to KDD Cup 2016. We used a two step approach in which we first identify full research papers corresponding to each conference of interest and then train two variants of exponential smoothing models to make predictions. Our solution achieves an overall score of 0.7508, while the winning submission scored 0.7656 in the overall results.

cs.AI↗

arXiv2015

Benchmarking KAZE and MCM for Multiclass Classification

In this paper, we propose a novel approach for feature generation by appropriately fusing KAZE and SIFT features. We then use this feature set along with Minimal Complexity Machine(MCM) for object classification. We show that KAZE and SIFT features are complementary. Experimental results indicate that an elementary integration of these techniques can outperform the state-of-the-art approaches.

cs.CV↗

arXiv2014

Improving Collaborative Filtering based Recommenders using Topic Modelling

Standard Collaborative Filtering (CF) algorithms make use of interactions between users and items in the form of implicit or explicit ratings alone for generating recommendations. Similarity among users or items is calculated purely based on rating overlap in this case,without considering explicit properties of users or items involved, limiting their applicability in domains with very sparse rating spaces. In many domains such as movies, news or electronic commerce recommenders, considerable contextual data in text form describing item properties is available along with the rating data, which could be utilized to improve recommendation quality.In this paper, we propose a novel approach to improve standard CF based recommenders by utilizing latent Dirichlet allocation (LDA) to learn latent properties of items, expressed in terms of topic proportions, derived from their textual description. We infer user's topic preferences or persona in the same latent space,based on her historical ratings. While computing similarity between users, we make use of a combined similarity measure involving rating overlap as well as similarity in the latent topic space. This approach alleviates sparsity problem as it allows calculation of similarity between users even if they have not rated any items in common. Our experiments on multiple public datasets indicate that the proposed hybrid approach significantly outperforms standard user Based and item Based CF recommenders in terms of classification accuracy metrics such as precision, recall and f-measure.

cs.IR↗