Showing 1–4 of 4 results for author: Paliwal, B

Visual Representations of Physiological Signals for Fake Video Detection

Authors: Kalin Stefanov, Bhawna Paliwal, Abhinav Dhall

Abstract: Realistic fake videos are a potential tool for spreading harmful misinformation given our increasing online presence and information intake. This paper presents a multimodal learning-based method for detection of real and fake videos. The method combines information from three modalities - audio, video, and physiology. We investigate two strategies for combining the video and physiology modalities… ▽ More Realistic fake videos are a potential tool for spreading harmful misinformation given our increasing online presence and information intake. This paper presents a multimodal learning-based method for detection of real and fake videos. The method combines information from three modalities - audio, video, and physiology. We investigate two strategies for combining the video and physiology modalities, either by augmenting the video with information from the physiology or by novelly learning the fusion of those two modalities with a proposed Graph Convolutional Network architecture. Both strategies for combining the two modalities rely on a novel method for generation of visual representations of physiological signals. The detection of real and fake videos is then based on the dissimilarity between the audio and modified video modalities. The proposed method is evaluated on two benchmark datasets and the results show significant increase in detection performance compared to previous methods. △ Less

Submitted 18 July, 2022; originally announced July 2022.

NGAME: Negative Mining-aware Mini-batching for Extreme Classification

Authors: Kunal Dahiya, Nilesh Gupta, Deepak Saini, Akshay Soni, Yajun Wang, Kushal Dave, Jian Jiao, Gururaj K, Prasenjit Dey, Amit Singh, Deepesh Hada, Vidit Jain, Bhawna Paliwal, Anshul Mittal, Sonu Mehta, Ramachandran Ramjee, Sumeet Agarwal, Purushottam Kar, Manik Varma

Abstract: Extreme Classification (XC) seeks to tag data points with the most relevant subset of labels from an extremely large label set. Performing deep XC with dense, learnt representations for data points and labels has attracted much attention due to its superiority over earlier XC methods that used sparse, hand-crafted features. Negative mining techniques have emerged as a critical component of all dee… ▽ More Extreme Classification (XC) seeks to tag data points with the most relevant subset of labels from an extremely large label set. Performing deep XC with dense, learnt representations for data points and labels has attracted much attention due to its superiority over earlier XC methods that used sparse, hand-crafted features. Negative mining techniques have emerged as a critical component of all deep XC methods that allow them to scale to millions of labels. However, despite recent advances, training deep XC models with large encoder architectures such as transformers remains challenging. This paper identifies that memory overheads of popular negative mining techniques often force mini-batch sizes to remain small and slow training down. In response, this paper introduces NGAME, a light-weight mini-batch creation technique that offers provably accurate in-batch negative samples. This allows training with larger mini-batches offering significantly faster convergence and higher accuracies than existing negative sampling techniques. NGAME was found to be up to 16% more accurate than state-of-the-art methods on a wide array of benchmark datasets for extreme classification, as well as 3% more accurate at retrieving search engine queries in response to a user webpage visit to show personalized ads. In live A/B tests on a popular search engine, NGAME yielded up to 23% gains in click-through-rates. △ Less

Submitted 10 July, 2022; originally announced July 2022.

Static IMRT and VMAT planning on 6 MV Flattened and Flattening-Filter-Free Beams of a TrueBeam VirtuaLinac System

Authors: Yue Yan, Poonam Yadav, Daniel Saenz, Bhudatt R. Paliwal

Abstract: In this manuscript, we investigated the accuracy of the entire chain of phase space files generated by the TrueBeam VirtuaLinac system in clinical treatment plans. Good agreement was obtained between the MC simulation and the clinical golden data for both flattened and FFF beams. The relative magnitude of agreement between MC and golden beam data presented in this research may assist a physicist i… ▽ More In this manuscript, we investigated the accuracy of the entire chain of phase space files generated by the TrueBeam VirtuaLinac system in clinical treatment plans. Good agreement was obtained between the MC simulation and the clinical golden data for both flattened and FFF beams. The relative magnitude of agreement between MC and golden beam data presented in this research may assist a physicist in terms of the percent dose deviation that one may expect when using MC data for verification purposes. △ Less

Submitted 23 September, 2015; originally announced September 2015.

Dosimetric and Biologic Differences in Flattened and Flattening-Filter-Free Beam Treatment Plans

Authors: Yue Yan, Poonam Yadav, Michael Bassetti, Kaifang Du, Daniel Saenz, Paul Harari, Bhudatt R. Paliwal

Abstract: Purpose: To quantitatively compare the dosimetric and biologic differences in treatment plans from flattened and flattening-filter-free (FFF) beam for three anatomic cancer sites. Methods and Materials: Treatment plans with static intensity-modulated radiotherapy beams and volumetric modulated arc therapy beams were generated for 13 patients for both the flattened beam and the FFF beam of the True… ▽ More Purpose: To quantitatively compare the dosimetric and biologic differences in treatment plans from flattened and flattening-filter-free (FFF) beam for three anatomic cancer sites. Methods and Materials: Treatment plans with static intensity-modulated radiotherapy beams and volumetric modulated arc therapy beams were generated for 13 patients for both the flattened beam and the FFF beam of the TrueBeam system. Beam energies of 6 MV and 10 MV were chosen for planning. A total of 104 treatment plans were generated in 13 patients. In order to analyze the biological effectiveness of treatment plans, dose volume histograms (DVH) were utilized. Flattened and FFF beam plans are quantitatively compared. Results: In head and neck cases, for VMAT plans, dose reduction in the FFF beam plans compared to the flattened beam in left cochlea, right submandibular gland and right parotid gland reached up to 2.36 Gy, 1.21 Gy and 1.45 Gy, respectively. Similarly, for static IMRT plans, the dose reduction of the FFF beam plans compared to the flattened beam plans for the same organs reached up to 0.34 Gy, 1.36 Gy and 1.46 Gy, respectively. Overall, for head and neck, the FFF beam plans achieved mean dose reduction of up to 5%, 7% and 9%, respectively for above organs at risk. For lung and prostate cases, the FFF beams provided lower or comparable NTCP values to organ-at-risk (OAR) compared to the flattened beam for all plans. Conclusions: In general, we observed treatment plans utilizing FFF beams can improve dose sparing to OARs without compromising the target coverage. Significant dose sparing effect is obtained for head and neck cancer cases, especially for the cases with relatively large field sizes (about 16x20 cm^2). For lung and prostate cases, compared to the flattened beam, the FFF beam based treatment plans provide lower or comparable dose to most OARs. △ Less

Submitted 16 June, 2015; originally announced June 2015.