2022 26th International Conference on Pattern Recognition (ICPR)
Handwritten document image binarization is challenging due to high variability in the written con... more Handwritten document image binarization is challenging due to high variability in the written content and complex background attributes such as page style, paper quality, stains, shadow gradients, and nonuniform illumination. While the traditional thresholding methods do not effectively generalize on such challenging real-world scenarios, deep learning-based methods have performed relatively well when provided with sufficient training data. However, the existing datasets are limited in size and diversity. This work proposes LS-HDIB-a large-scale handwritten document image binarization dataset containing over a million document images that span numerous realworld scenarios. Additionally, we introduce a novel technique that uses a combination of adaptive thresholding and seamless cloning methods to create the dataset with accurate ground truths. Through an extensive quantitative and qualitative evaluation over eight different deep learning based models, we demonstrate the enhancement in the performance of these models when trained on the LS-HDIB dataset and tested on unseen images.
Automatic extraction of raw data from 2D line plot images is a problem of great importance having... more Automatic extraction of raw data from 2D line plot images is a problem of great importance having many real-world applications. Several algorithms have been proposed for solving this problem. However, these algorithms involve a significant amount of human intervention. To minimize this intervention, we propose APEX-Net, a deep learning based framework with novel loss functions for solving the plot extraction problem. We introduce APEX-1M, a new large scale dataset which contains both the plot images and the raw data. We demonstrate the performance of APEX-Net on the APEX-1M test set and show that it obtains impressive accuracy. We also show visual results of our network on unseen plot images and demonstrate that it extracts the shape of the plots to a great extent. Finally, we develop a GUI based software for plot extraction that can benefit the community at large. For dataset and more information visit https://sites. google.com/view/apexnetpaper/.
2022 IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW)
We consider the generic deep image enhancement problem where an input image is transformed into a... more We consider the generic deep image enhancement problem where an input image is transformed into a perceptually better-looking image. Recent methods for image enhancement consider the problem by performing style transfer and image restoration. The methods mostly fall into two categories: training data-based and training data-independent (deep internal learning methods). We perform image enhancement in the deep internal learning framework. Our Deep Internal Learning for Image Enhancement framework enhances content features and style features and uses contextual content loss for preserving image context in the enhanced image. We show results on both hazy and noisy image enhancement. To validate the results, we use structure similarity and perceptual error, which is efficient in measuring the unrealistic deformation present in the images. We show that the proposed framework outperforms the relevant state-of-the-art works for image enhancement.
Handwritten document image binarization is a challenging task due to high diversity in the conten... more Handwritten document image binarization is a challenging task due to high diversity in the content, page style, and condition of the documents. While the traditional thresholding methods fail to generalize on such challenging scenarios, deep learning based methods can generalize well however, require a large training data. Current datasets for handwritten document image binarization are limited in size and fail to represent several challenging real-world scenarios. To solve this problem, we propose HDIB1M a handwritten document image binarization dataset of 1M images. We also present a novel method used to generate this dataset. To show the effectiveness of our dataset we train a deep learning model UNetED on our dataset and evaluate its performance on other publicly available datasets. The dataset and the code will be made available to the community.
Transport through 2D correlated porous media has been shown to induce mixing through deformation ... more Transport through 2D correlated porous media has been shown to induce mixing through deformation of material elements owing to the heterogeneity in the flow fields [1]. The stretching of line elements brings about an increase in the area for reaction and increases the diffusive flux through the thinning of the direction transverse to stretching and is thought to be an important mechanism to determine hotspots of mixing in various flow conditions [2]. These mechanisms are useful to quantify the extent of reaction for fast reactive fluids wherein the local reactivity may be obtained through the information gained from the gradient maps of a conservative tracer. Unlike earlier works which have primarily focussed on a Lagrangian approach for quantifying dispersion, we attempt to establish the dispersion rates in the transport through a porous media by means of fully resolved Eulerian simulations for the typical Peclet numbers (ratio of the diffusion time to the advection time) encounter...
Point cloud is an efficient way of representing and storing 3D geometric data. Deep learning algo... more Point cloud is an efficient way of representing and storing 3D geometric data. Deep learning algorithms on point clouds are time and memory efficient. Several methods such as PointNet and FoldingNet have been proposed for processing point clouds. This work proposes an autoencoder based framework to generate robust embeddings for point clouds by utilizing hierarchical information using graph convolution. We perform multiple experiments to assess the quality of embeddings generated by the proposed encoder architecture and visualize the t-SNE map to highlight its ability to distinguish between different object classes. We further demonstrate the applicability of the proposed framework in applications like: 3D point cloud completion and Single image based 3D reconstruction.
We analyze the dynamics of solute mixing and reaction in a mixing-limited reactive flow by consid... more We analyze the dynamics of solute mixing and reaction in a mixing-limited reactive flow by considering the transport of a tracer in a linear shear flow and in a Rankine vortex. The action of a shear flow, in general, achieves stretching of fluid elements due to the heterogeneous nature of the flow. A vortex flow exhibits not only stretching but also folding of fluid elements in a way that brings adjacent fluid elements closer at every turn. A strong stretching along the tangential direction is accompanied by a concomitant thinning in the radial direction leading to a strong diffusive flux, which may cause the material from neighboring regions of the mixing interface to aggregate. Through a Lagrangian concentration evolution technique, the diffusive strip method, we obtain the concentration field and pinpoint the signature of coalescence of two neighboring concentration regions by analyzing the concentration distribution profiles. The role of substrate deformation on the reaction kinetics of a classical heterogeneous chemical reaction is also studied where we derive analytical expressions for the coupling between the rate of product formation and the Péclet number in different time limits. Finally, the impact of coalescence on reaction rates is studied for a Rankine vortex, a result that holds important implications for simple bimolecular reactions. This analysis is useful to understand scalar dispersion in vortical flow structures and the consequences of stretching-enhanced diffusion in mixing-limited reactive flows.
2022 26th International Conference on Pattern Recognition (ICPR)
Handwritten document image binarization is challenging due to high variability in the written con... more Handwritten document image binarization is challenging due to high variability in the written content and complex background attributes such as page style, paper quality, stains, shadow gradients, and nonuniform illumination. While the traditional thresholding methods do not effectively generalize on such challenging real-world scenarios, deep learning-based methods have performed relatively well when provided with sufficient training data. However, the existing datasets are limited in size and diversity. This work proposes LS-HDIB-a large-scale handwritten document image binarization dataset containing over a million document images that span numerous realworld scenarios. Additionally, we introduce a novel technique that uses a combination of adaptive thresholding and seamless cloning methods to create the dataset with accurate ground truths. Through an extensive quantitative and qualitative evaluation over eight different deep learning based models, we demonstrate the enhancement in the performance of these models when trained on the LS-HDIB dataset and tested on unseen images.
Automatic extraction of raw data from 2D line plot images is a problem of great importance having... more Automatic extraction of raw data from 2D line plot images is a problem of great importance having many real-world applications. Several algorithms have been proposed for solving this problem. However, these algorithms involve a significant amount of human intervention. To minimize this intervention, we propose APEX-Net, a deep learning based framework with novel loss functions for solving the plot extraction problem. We introduce APEX-1M, a new large scale dataset which contains both the plot images and the raw data. We demonstrate the performance of APEX-Net on the APEX-1M test set and show that it obtains impressive accuracy. We also show visual results of our network on unseen plot images and demonstrate that it extracts the shape of the plots to a great extent. Finally, we develop a GUI based software for plot extraction that can benefit the community at large. For dataset and more information visit https://sites. google.com/view/apexnetpaper/.
2022 IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW)
We consider the generic deep image enhancement problem where an input image is transformed into a... more We consider the generic deep image enhancement problem where an input image is transformed into a perceptually better-looking image. Recent methods for image enhancement consider the problem by performing style transfer and image restoration. The methods mostly fall into two categories: training data-based and training data-independent (deep internal learning methods). We perform image enhancement in the deep internal learning framework. Our Deep Internal Learning for Image Enhancement framework enhances content features and style features and uses contextual content loss for preserving image context in the enhanced image. We show results on both hazy and noisy image enhancement. To validate the results, we use structure similarity and perceptual error, which is efficient in measuring the unrealistic deformation present in the images. We show that the proposed framework outperforms the relevant state-of-the-art works for image enhancement.
Handwritten document image binarization is a challenging task due to high diversity in the conten... more Handwritten document image binarization is a challenging task due to high diversity in the content, page style, and condition of the documents. While the traditional thresholding methods fail to generalize on such challenging scenarios, deep learning based methods can generalize well however, require a large training data. Current datasets for handwritten document image binarization are limited in size and fail to represent several challenging real-world scenarios. To solve this problem, we propose HDIB1M a handwritten document image binarization dataset of 1M images. We also present a novel method used to generate this dataset. To show the effectiveness of our dataset we train a deep learning model UNetED on our dataset and evaluate its performance on other publicly available datasets. The dataset and the code will be made available to the community.
Transport through 2D correlated porous media has been shown to induce mixing through deformation ... more Transport through 2D correlated porous media has been shown to induce mixing through deformation of material elements owing to the heterogeneity in the flow fields [1]. The stretching of line elements brings about an increase in the area for reaction and increases the diffusive flux through the thinning of the direction transverse to stretching and is thought to be an important mechanism to determine hotspots of mixing in various flow conditions [2]. These mechanisms are useful to quantify the extent of reaction for fast reactive fluids wherein the local reactivity may be obtained through the information gained from the gradient maps of a conservative tracer. Unlike earlier works which have primarily focussed on a Lagrangian approach for quantifying dispersion, we attempt to establish the dispersion rates in the transport through a porous media by means of fully resolved Eulerian simulations for the typical Peclet numbers (ratio of the diffusion time to the advection time) encounter...
Point cloud is an efficient way of representing and storing 3D geometric data. Deep learning algo... more Point cloud is an efficient way of representing and storing 3D geometric data. Deep learning algorithms on point clouds are time and memory efficient. Several methods such as PointNet and FoldingNet have been proposed for processing point clouds. This work proposes an autoencoder based framework to generate robust embeddings for point clouds by utilizing hierarchical information using graph convolution. We perform multiple experiments to assess the quality of embeddings generated by the proposed encoder architecture and visualize the t-SNE map to highlight its ability to distinguish between different object classes. We further demonstrate the applicability of the proposed framework in applications like: 3D point cloud completion and Single image based 3D reconstruction.
We analyze the dynamics of solute mixing and reaction in a mixing-limited reactive flow by consid... more We analyze the dynamics of solute mixing and reaction in a mixing-limited reactive flow by considering the transport of a tracer in a linear shear flow and in a Rankine vortex. The action of a shear flow, in general, achieves stretching of fluid elements due to the heterogeneous nature of the flow. A vortex flow exhibits not only stretching but also folding of fluid elements in a way that brings adjacent fluid elements closer at every turn. A strong stretching along the tangential direction is accompanied by a concomitant thinning in the radial direction leading to a strong diffusive flux, which may cause the material from neighboring regions of the mixing interface to aggregate. Through a Lagrangian concentration evolution technique, the diffusive strip method, we obtain the concentration field and pinpoint the signature of coalescence of two neighboring concentration regions by analyzing the concentration distribution profiles. The role of substrate deformation on the reaction kinetics of a classical heterogeneous chemical reaction is also studied where we derive analytical expressions for the coupling between the rate of product formation and the Péclet number in different time limits. Finally, the impact of coalescence on reaction rates is studied for a Rankine vortex, a result that holds important implications for simple bimolecular reactions. This analysis is useful to understand scalar dispersion in vortical flow structures and the consequences of stretching-enhanced diffusion in mixing-limited reactive flows.
Uploads
Papers by prajwal singh