Meta tags:
Headings (most frequently used words):
saurabh, saini, about, me, publications, current, projects,
Text of the page (most frequently used words):
and (119), the (113), for (49), image (46), #saurabh (31), saini (31), our (30), narayanan (27), with (26), this (24), using (23), method (22), iid (21), scene (19), priors (19), that (18), semantic (16), can (15), abstract (15), computer (15), decomposition (15), author (14), title (14), year (14), bibtex (14), vision (14), from (13), video (13), arxiv (13), intrinsic (13), low (12), conference (12), 2016 (12), use (12), icvgip (12), neural (12), concept (12), light (11), network (11), inproceedings (11), booktitle (11), paper (11), tags (11), learning (11), images (11), novel (11), results (11), 2018 (11), enhancement (10), doi (10), methods (10), model (10), features (9), information (9), hash (9), training (9), tracking (9), webpage (9), proceedings (8), present (8), show (8), oral (8), are (8), code (8), human (8), capture (8), fields (8), context (8), exposure (8), illumination (7), based (7), which (7), problem (7), space (7), processing (7), 1145 (7), state (7), art (7), view (7), material (7), poster (7), into (7), fusion (7), dblp (6), singh (6), pages (6), url (6), like (6), datasets (6), end (6), representation (6), graphics (6), focal (6), scenes (6), captured (6), body (6), propose (6), single (6), point (6), several (6), hierarchical (6), level (6), prior (6), research (6), views (6), radiance (6), 2022 (6), appearance (6), multiple (6), concepts (6), segmentation (6), conf (5), traditional (5), domain (5), short (5), clips (5), https (5), embedding (5), tag2vec (5), indian (5), visual (5), two (5), system (5), stacks (5), optimization (5), objects (5), also (5), supplementary (5), applications (5), analysis (5), other (5), stylization (5), stylized (5), factorization (5), under (4), techniques (4), current (4), aditya (4), rajvi (4), shah (4), adaptation (4), classification (4), org (4), both (4), tag (4), 3009977 (4), given (4), direct (4), step (4), parikshit (4), sakurikar (4), stack (4), object (4), has (4), these (4), used (4), local (4), global (4), pixel (4), lighting (4), pdf (4), wacv (4), ieee (4), challenging (4), evaluation (4), involves (4), each (4), knowledge (4), visualization (4), 2019 (4), rendering (4), work (4), over (4), better (4), interactive (4), consistent (4), cloud (4), goel (4), rahul (4), sirikonda (4), dhawal (4), gupta (4), avani (4), components (4), quaternion (4), simulated (4), 2023 (4), cvpr (4), factors (4), dot (4), decomposing (3), generating (3), varying (3), maps (3), robust (3), leveraging (3), defense (3), against (3), action (3), social (3), pattern (3), recognition (3), vines (3), wild (3), compared (3), focus (3), simple (3), source (3), target (3), set (3), performance (3), videos (3), india (3), understanding (3), such (3), new (3), have (3), relevant (3), useful (3), approaches (3), learn (3), first (3), natural (3), models (3), dimensional (3), then (3), without (3), any (3), directly (3), rgbf (3), input (3), rgbd (3), highlight (3), sparse (3), structure (3), correspondences (3), standard (3), generated (3), competitive (3), shape (3), typically (3), part (3), arbitrary (3), subject (3), not (3), feature (3), cmesh (3), bmvc (3), analyzing (3), flexible (3), regions (3), search (3), convolutional (3), framework (3), finally (3), produce (3), deep (3), article (3), journal (3), materials (3), iterative (3), quality (3), high (3), scheme (3), shading (3), reflectance (3), stylize (3), 2021 (3), cvf (3), style (3), synthesis (3), matrix (3), styletrf (3), 3571600 (3), thirteenth (3), series (3), problems (3), etc (3), interpretability (3), sensitivity (3), factorized (3), layers (3), distillation (3), centered (3), explanations (3), specularity (3), lle (3), project (2), spectra (2), spatially (2), submission (2), unrolled (2), architecture (2), employed (2), interpretable (2), backdoor (2), singhssn16 (2), modern (2), gcpr (2), september (2), 1007 (2), 978 (2), 319 (2), 45886 (2), _20 (2), data (2), strategy (2), utilize (2), labeled (2), iteratively (2), function (2), additionally (2), available (2), technique (2), test (2), tenth (2), guwahati (2), assam (2), december (2), 3010035 (2), user (2), media (2), recently (2), known (2), popular (2), study (2), driven (2), transfer (2), vector (2), train (2), nearest (2), neighbor (2), validate (2), 3010046 (2), scale (2), beyond (2), rgb (2), towards (2), depth (2), measure (2), random (2), dense (2), across (2), neighbourhood (2), well (2), dataset (2), provides (2), gaurav (2), mishra (2), kiran (2), varanasi (2), home (2), requires (2), kinect (2), detection (2), motion (2), channel (2), when (2), camera (2), consists (2), statistical (2), along (2), improve (2), machine (2), interesting (2), various (2), integrated (2), approach (2), three (2), levels (2), properties (2), within (2), small (2), mid (2), encode (2), semantics (2), establish (2), designed (2), fixed (2), selective (2), experiments (2), indicate (2), utility (2), structured (2), compare (2), lesser (2), artifacts (2), proper (2), choice (2), encoding (2), signifying (2), importance (2), believe (2), insights (2), presented (2), would (2), future (2), best (2), aakash (2), renderer (2), artist (2), real (2), world (2), workflow (2), modeling (2), editing (2), time (2), selection (2), appropriate (2), generates (2), control (2), environment (2), more (2), rendered (2), comparison (2), faster (2), tool (2), sparsity (2), contemporary (2), hsin (2), ping (2), huang (2), hung (2), tseng (2), maneesh (2), ming (2), hsuan (2), yang (2), international (2), iccv (2), reference (2), different (2), diverse (2), stylizing (2), tensorial (2), generation (2), casually (2), much (2), previous (2), its (2), manner (2), require (2), joint (2), tensorf (2), few (2), interpreting (2), activations (2), invariant (2), ground (2), truth (2), specific (2), metrics (2), disentanglement (2), order (2), specifically (2), factorizing (2), specular (2), task (2), rfs (2), important (2), fine (2), grows (2), improvement (2), systems (2), networks (2), cavs (2), estimate (2), biases (2), hoc (2), reduce (2), layer (2), class (2), sensitive (2), induce (2), reconstruction (2), 2024 (2), property (2), tasks (2), related (2), technology (2), iiit (2), center (2), aims, studying, feasibility, color, illuminanat, illuination, relit, separation, build, retnex, self, supervised, retinex, attacks, conceptual, projects, dagm, 38th, german, hannover, germany, 245, 257, internet, significantly, distribution, unsupervised, beled, augmentation, word2vec, common, subspace, embed, unlabled, incrementally, augments, samples, modifies, bring, distributions, together, multi, modal, incorporates, noisy, form, effectiveness, achieve, notable, improvements, labels, valuable, resources, type, labeling, mechanism, become, increasingly, sites, enrichment, recommendation, similarity, label, propagation, attempt, cost, mapping, process, employ, language, nlp, skip, gram, corpus, million, map, categories, query, mapped, learned, found, performing, retrieval, relevance, suggested, qualitatively, quantitatively, sainisn16, presents, restrictions, complexity, captures, distances, since, depends, distance, call, generalized, walk, algorithm, compute, probability, define, neighbourhoods, adhering, underlying, chromaticity, assumptions, constraints, indoor, outdoor, manually, artificial, conditions, larger, synthetically, nyuv2, mpi, sintel, strong, evidence, potential, modality, place, mishrasvn18, winter, lake, tahoe, usa, march, 390, 399, 1109, 00049, specialized, ups, although, pose, consumer, devices, microsoft, restricted, stick, figures, visualizing, full, movements, wears, casual, clothes, initialization, procedure, move, around, front, makes, applicable, clothing, textures, environments, minimal, intervention, surface, articulated, regularized, idea, consensus, mesh, template, person, created, demonstrate, poses, argue, accuracies, quantitative, shows, very, little, drift, improved, usage, presentation, sainin18, british, northumbria, university, newcastle, 154, http, bmvc2018, contents, papers, 0796, neighborhood, weak, industrial, honourable, mention, 2019arxiv190809530k, prints, keywords, science, month, aug, archiveprefix, eprint, 1908, 09530, primaryclass, photo, realism, imagery, crucially, dependent, how, able, recreate, manual, tweaking, parameters, uses, path, tracing, engine, feedback, lot, may, spent, quickly, ray, traced, visualizations, shaderball, allows, assists, ability, render, enables, gives, freedom, denoising, suggests, performs, provide, release, foster, further, area, siggraph, asia, technical, briefs, saini2019semantichp, volume, abs, 1902, 03830, multistage, complementary, formulations, smoothing, solutions, even, huang2021learningts, kumar, 13849, 13858, tackle, same, desired, inputs, solution, combining, lead, blurry, construct, back, projecting, second, develop, aggregation, modules, gather, modulate, linear, transformation, transformed, obtain, experimental, alternative, goel2022styletrf, 3571643, received, attention, geometry, sets, jointly, snerf, trains, nerf, alternating, compact, quick, optimize, finetuned, iterations, thus, effectively, decouples, adaption, than, purpose, spotlight, guptaiid22, 3571603, ill, posed, constituent, albedo, large, carried, out, either, synthetic, sparsely, annotated, split, valid, ways, one, lmse, mse, dssim, whdr, saw, inadequate, measuring, way, evaluate, inspired, csm, award, sainiqfsef22, 3571604, maximizes, location, merging, content, enhanced, fuses, bracketed, poorly, lit, generate, properly, illuminated, still, leading, contrast, qfsef, representing, pure, theoretically, dichromatic, reflection, accounts, characteristics, progressively, separate, diffuse, empirically, prove, advantages, simulation, isrfgoel2023, represent, mixed, reality, personal, spaces, needs, manipulating, represented, efforts, promise, but, complex, isrf, interactively, segment, matching, distilled, identifies, confidence, seed, bilateral, spatio, region, recover, accurate, segmenting, compositing, them, another, changing, others, gupta2023concept, thirty, seventh, openreview, net, forum, arkmhtyll6, humans, instead, hard, recent, focused, activation, vectors, possible, extend, post, ante, bias, through, tuning, additional, loss, were, defined, final, past, generalize, intermediate, prototypes, facilitates, last, convolution, most, informative, introduce, create, richer, pre, trained, knowledgeable, teacher, sensitize, desensitize, debias, please, visit, details, neurips, saini2024rsfnet, enhance, insufficient, been, proposed, years, many, additive, treats, composed, latent, estimated, modulating, during, learnable, rsfnet, unrolling, design, manipulated, zero, need, paired, unpaired, supervision, improves, benchmarks, achieves, generalization, supplement, inducing, deraining, deblurring, dehazing, negligible, overhead, shown, publications, finishing, pending, guidance, lab, general, interest, topics, especially, applied, currently, working, focusing, separating, relighting, apart, wandering, domains, registration, geometric, explainable, list, personally, sometimes, enjoy, sketching, graphite, reading, non, fiction, writing, poems, inverse, institute, hyderabad, kohli, intelligent, kcis, cvit, prof, about, google, scholar, linkedin,
Text of the page (random words):
s focused on human centered concept explanations of neural networks concept activation vectors cavs estimate a model s sensitivity and possible biases to a given concept in this paper we extend cavs from post hoc analysis to ante hoc training in order to reduce model bias through fine tuning using an additional concept loss concepts were defined on the final layer of the network in the past we generalize it to intermediate layers using class prototypes this facilitates class learning in the last convolution layer which is known to be most informative we also introduce concept distillation to create richer concepts using a pre trained knowledgeable model as the teacher our method can sensitize or desensitize a model towards concepts we show applications of concept sensitive training to debias several classification problems we also use concepts to induce prior knowledge into iid a reconstruction problem concept sensitive training can improve model interpretability reduce biases and induce prior knowledge please visit this https url for code and more details inproceedings gupta2023concept title concept distillation leveraging human centered explanations for model improvement author avani gupta and saurabh saini and p j narayanan booktitle thirty seventh conference on neural information processing systems year 2023 url https openreview net forum id arkmhtyll6 interactive segmentation of radiance fields rahul goel sirikonda dhawal saurabh saini p j narayanan cvpr 2023 poster abstract bibtex webpage code arxiv video radiance fields rf are popular to represent casually captured scenes for new view synthesis and several applications beyond it mixed reality on personal spaces needs understanding and manipulating scenes represented as rfs with semantic segmentation of objects as an important step prior segmentation efforts show promise but do not scale to complex objects with diverse appearance we present the isrf method to interactively segment objects with fine structure and appearance nearest neighbor feature matching using distilled semantic features identifies high confidence seed regions bilateral search in a joint spatio semantic space grows the region to recover accurate segmentation we show state of the art results of segmenting objects from rfs and compositing them to another scene changing appearance etc and an interactive segmentation tool that others can use inproceedings isrfgoel2023 title interactive segmentation of radiance fields author goel rahul and sirikonda dhawal and saini saurabh and narayanan p j year 2023 booktitle proceedings of the ieee cvf conference on computer vision and pattern recognition cvpr quaternion factorized simulated exposure fusion for low light image enhancement saurabh saini p j narayanan icvgip 2022 oral abstract bibtex webpage code poster image fusion maximizes the visual information at each pixel location by merging content from multiple images in order to produce an enhanced image exposure fusion specifically fuses a bracketed exposure stack of poorly lit images to generate a properly illuminated image given a single input image exposure fusion can still be employed on a simulated exposure stack leading to direct single image contrast and low light enhancement in this work we present a novel quaternion factorized simulated exposure fusion qfsef method by factorizing an input image into multiple illumination consistent layers to this end we use an iterative sparse matrix factorization scheme by representing the image as a two dimensional pure quaternion matrix theoretically our representation is based on the dichromatic reflection model and accounts for the two scene illumination characteristics by factorizing each progressively generated image into separate specular and diffuse components we empirically prove the advantages of our factorization scheme over other exposure simulation methods by using it for the low light image enhancement task inproceedings sainiqfsef22 author saini saurabh and narayanan p j title quaternion factorized simulated exposure fusion for low light image enhancement year 2022 doi 10 1145 3571600 3571604 booktitle proceedings of the thirteenth indian conference on computer vision graphics and image processing series icvgip 22 interpreting intrinsic image decomposition using concept activations avani gupta saurabh saini p j narayanan icvgip 2022 oral best paper award abstract bibtex webpage code poster supplementary evaluation of ill posed problems like intrinsic image decomposition iid is challenging iid involves decomposing an image into its constituent illumination invariant reflectance r and albedo invariant shading s components contemporary iid methods use deep learning models and require large datasets for training the evaluation of iid is carried out on either synthetic ground truth images or sparsely annotated natural images a scene can be split into reflectance and shading in multiple valid ways comparison with one specific decomposition in the ground truth images used by current iid evaluation metrics like lmse mse dssim whdr saw ap etc is inadequate measuring r s disentanglement is a better way to evaluate the quality of iid inspired by ml interpretability methods we propose concept sensitivity metrics csm that directly measure disentanglement using sensitivity to relevant concepts inproceedings guptaiid22 author gupta avani and saini saurabh and narayanan p j title interpreting intrinsic image decomposition using concept activations year 2022 doi 10 1145 3571600 3571603 booktitle proceedings of the thirteenth indian conference on computer vision graphics and image processing series icvgip 22 styletrf stylizing tensorial radiance fields rahul goel sirikonda dhawal saurabh saini p j narayanan icvgip 2022 spotlight abstract bibtex webpage arxiv poster video stylized view generation of scenes captured casually using a camera has received much attention recently the geometry and appearance of the scene are typically captured as neural point sets or neural radiance fields in the previous work an image stylization method is used to stylize the captured appearance by training its network jointly or iteratively with the structure capture network the state of the art snerf method trains the nerf and stylization network in an alternating manner these methods have high training time and require joint optimization in this work we present styletrf a compact quick to optimize strategy for stylized view generation using tensorf the appearance part is finetuned using sparse stylized priors of a few views rendered using the tensorf representation for a few iterations our method thus effectively decouples style adaption from view capture and is much faster than the previous methods we show state of the art results on several scenes used for this purpose inproceedings goel2022styletrf author goel rahul and sirikonda dhawal and saini saurabh and narayanan p j title styletrf stylizing tensorial radiance fields year 2022 doi 10 1145 3571600 3571643 booktitle proceedings of the thirteenth indian conference on computer vision graphics and image processing series icvgip 22 learning to stylize novel views hsin ping huang hung yu tseng saurabh saini maneesh singh ming hsuan yang iccv 2021 poster abstract bibtex webpage code arxiv supplementary we tackle a 3d scene stylization problem generating stylized images of a scene from arbitrary novel views given a set of images of the same scene and a reference image of the desired style as inputs direct solution of combining novel view synthesis and stylization approaches lead to results that are blurry or not consistent across different views we propose a point cloud based method for consistent 3d scene stylization first we construct the point cloud by back projecting the image features to the 3d space second we develop point cloud aggregation modules to gather the style information of the 3d scene and then modulate the features in the point cloud with a linear transformation matrix finally we project the transformed features to 2d space to obtain the novel views experimental results on two diverse datasets of real world scenes validate that our method generates consistent stylized novel view synthesis results against other alternative approaches article huang2021learningts title learning to stylize novel views author hsin ping huang and hung yu tseng and saurabh saini and maneesh kumar singh and ming hsuan yang journal 2021 ieee cvf international conference on computer vision iccv year 2021 pages 13849 13858 semantic hierarchical priors for intrinsic image decomposition saurabh saini p j narayanan arxiv 2019 abstract bibtex arxiv intrinsic image decomposition iid is a challenging and interesting computer vision problem with various applications in several fields we present novel semantic priors and an integrated approach for single image iid that involves analyzing image at three hierarchical context levels local context priors capture scene properties at each pixel within a small neighbourhood mid level context priors encode object level semantics global context priors establish correspondences at the scene level our semantic priors are designed on both fixed and flexible regions using selective search method and convolutional neural network features our iid method is an iterative multistage optimization scheme and consists of two complementary formulations l2 smoothing for shading and l1 sparsity for reflectance experiments and analysis of our method indicate the utility of our semantic priors and structured hierarchical analysis in an iid framework we compare our method with other contemporary iid solutions and show results with lesser artifacts finally we highlight that proper choice and encoding of prior knowledge can produce competitive results even when compared to end to end deep learning iid methods signifying the importance of such priors we believe that the insights and techniques presented in this paper would be useful in the future iid research article saini2019semantichp title semantic hierarchical priors for intrinsic image decomposition author saurabh saini and p j narayanan journal arxiv year 2019 volume abs 1902 03830 neural rendering for material visualization aakash kt parikshit sakurikar saurabh saini p j narayanan siggraph asia 2019 technical briefs abstract bibtex webpage code arxiv photo realism in computer generated imagery is crucially dependent on how well an artist is able to recreate real world materials in the scene the workflow for material modeling and editing typically involves manual tweaking of material parameters and uses a standard path tracing engine for visual feedback a lot of time may be spent in iterative selection and rendering of materials at an appropriate quality in this work we propose a convolutional neural network based workflow which quickly generates high quality ray traced material visualizations on a shaderball our novel architecture allows for control over environment lighting and assists material selection along with the ability to render spatially varying materials additionally our network enables control over environment lighting which gives an artist more freedom and provides better visualization of the rendered material comparison with state of the art denoising and neural rendering techniques suggests that our neural renderer performs faster and better we provide a interactive visualization tool and release our training dataset to foster further research in this area article 2019arxiv190809530k author kt aakash and sakurikar parikshit and saini saurabh and narayanan p j title a flexible neural renderer for material visualization journal arxiv e prints keywords computer science graphics year 2019 month aug archiveprefix arxiv eprint 1908 09530 primaryclass cs gr semantic priors for intrinsic image decomposition saurabh saini p j narayanan bmvc 2018 oral best industrial paper honourable mention abstract bibtex pdf supplementary intrinsic image decomposition iid is a challenging and interesting computer vision problem with various applications in several fields we present novel semantic priors and an integrated approach for single image iid that involves analyzing image at three hierarchical context levels local context priors capture scene properties at each pixel within a small neighborhood mid level context priors encode object level semantics global context priors establish correspondences at the scene level our semantic priors are designed on both fixed and flexible regions using selective search method and convolutional neural network features experiments and analysis of our method indicate the utility of our weak semantic priors and structured hierarchical analysis in an iid framework we compare our method with the current state of the art and show results with lesser artifacts finally we highlight that proper choice and encoding of prior knowledge can produce competitive results compared to end to end deep learning iid methods signifying the importance of such priors we believe that the insights and techniques presented in this paper would be useful in the future iid research inproceedings dblp conf bmvc sainin18 author saurabh saini and p j narayanan title semantic priors for intrinsic image decomposition booktitle british machine vision conference 2018 bmvc 2018 northumbria university newcastle uk september 3 6 2018 pages 154 year 2018 url http bmvc2018 org contents papers 0796 pdf human shape capture and tracking at home gaurav mishra saurabh saini kiran varanasi p j narayanan wacv 2018 oral abstract bibtex webpage pdf supplementary presentation human body tracking typically requires specialized capture set ups although pose tracking is available in consumer devices like microsoft kinect it is restricted to stick figures visualizing body part detection in this paper we propose a method for full 3d human body shape and motion capture of arbitrary movements from the depth channel of a single kinect when the subject wears casual clothes we do not use the rgb channel or an initialization procedure that requires the subject to move around in front of the camera this makes our method applicable for arbitrary clothing textures and lighting environments with minimal subject intervention our method consists of 3d surface feature detection and articulated motion tracking which is regularized by a statistical human body model we also propose the idea of a consensus mesh cmesh which is the 3d template of a person created from a single view point we demonstrate tracking results on challenging poses and argue that using cmesh along with statistical body models can improve tracking accuracies quantitative evaluation of our dense body tracking shows that our method has very little drift which is improved by the usage of cmesh inproceedings dblp conf wacv mishrasvn18 author gaurav mishra and saurabh saini and kiran varanasi and p j narayanan title human shape capture and t...
|
