1

Joint Video Rolling Shutter Correction and Super-Resolution

We propose PatchNet that bi-directional motion field in the feature space to extract relevant information from neighboring patches using attention mechanism, and deformable fields using deformable convolutions to extract local pixel-level information for joint rolling shutter correction and super-resolution.

GAMA: Generative Adversarial Multi-Object Scene Attacks

GAMA demonstrates the utility of the CLIP model as an attacker’s tool to train formidable perturbation generators for multi-object scenes. Using the joint image-text features to train the generator, we show that GAMA can craft potent transferable perturbations in order to fool victim classifiers in various attack settings.

Poisson2Sparse: Self-Supervised Poisson Denoising From a Single Image

We propose a self-supervised learning methos that utilizes sparsity and dictionary learning-based approach for single-image denoising where the noise is approximated as a Poisson process, requiring no clean ground-truth data.

Ada-VSR: Adaptive Video Super-Resolution with Meta-Learning

We propose Ada-VSR framework that employs meta-learning to obtain adaptive model parameters, using a large-scale external dataset, that can adapt quickly to novel conditions of the given test video, thereby exploiting external and internal information of a video for super-resolution.

ALANET: Adaptive Latent Attention Network for Joint Video Deblurring and Interpolation

We introduce a novel architecture, Adaptive Latent Attention Network (ALANET), which synthesizes sharp high frame-rate videos with no prior knowledge of input frames being blurry or not, thereby performing the task of both deblurring and interpolation.

Adversarial Knowledge Transfer from Unlabeled Data

We present a novel adversarial framework for transferring knowledge from internet-scale unlabeled data to improve the performance of a classifier on a given visual recognition task.

Non-Adversarial Video Synthesis with Learned Priors

We present a non-adversarial approach for video synthesis by jointly optimizing the latent space and the model parametres.

Deep Quantized Representation for Enhanced Reconstruction

We propose a data-driven Deep Quantized Latent Representation (DQLR) methodology for high-quality image reconstruction in the Shoot Apical Meristem (SAM) of Arabidopsis thaliana by utilizing multiple consecutive slices in the z-stack

Deep Learning based Identity Verification in Renaissance Portraits

We present a Siamese Convolutional Neural Networks (CNN) based deep learning algorithm for the task of face recognition in renaissance portraits.