arXiv subjects

Vijayakrishna Naganoor

Publications and source records attributed to Vijayakrishna Naganoor.

3 recordsLinked to original sources

arXiv2026

IRIS: A Visual Cortex-Inspired Framework for Analyzing Orientation Selectivity in Vision Transformers

Vision transformers (ViTs) have become the de facto standard for image encoding across many perception tasks. Despite their empirical success, it remains mechanistically unclear how they encode low-level features, given their lack of inductive biases: ViTs process information globally rather than relying on local structure. Biological visual systems, in contrast, build low-level features, such as orientation selectivity in the primary visual cortex, by combining information from small, localized regions of the visual field. These features are general-purpose representations, shared and required across multiple specialized neural pathways, unlike higher-level, task-specific semantic features. This raises the question if such biologically-grounded features arise in ViTs. In this work, we systematically study how orientation selectivity emerges in ViTs by introducing a suite of neuroscience-inspired metrics: representational similarity score (RSS), orientation recruitment score (ORS), and orientation tuning bandwidth to quantify how orientation is encoded in representational geometry and as a function of model depth. Through extensive analysis, we find that: (1) the training paradigm is the strongest determinant of orientation selectivity, with models sharing an objective, peaking at comparable relative depths regardless of scale (2) many units are orientation-selective early in training, with early-to-middle layers recruiting more such units over time, while deeper layers lose selectivity and broaden their tuning toward semantic encoding and (3) our metrics offer a mechanistic heuristic for how many layers to unfreeze for best downstream generalization. Our framework presents a way to track biologically-grounded features during ViT training, probes how desired properties are encoded in transformer representations, and builds a systematic understanding of how ViTs generalize across tasks.

cs.CV↗

arXiv2020

CompressNet: Generative Compression at Extremely Low Bitrates

Compressing images at extremely low bitrates (< 0.1 bpp) has always been a challenging task since the quality of reconstruction significantly reduces due to the strong imposed constraint on the number of bits allocated for the compressed data. With the increasing need to transfer large amounts of images with limited bandwidth, compressing images to very low sizes is a crucial task. However, the existing methods are not effective at extremely low bitrates. To address this need, we propose a novel network called CompressNet which augments a Stacked Autoencoder with a Switch Prediction Network (SAE-SPN). This helps in the reconstruction of visually pleasing images at these low bitrates (< 0.1 bpp). We benchmark the performance of our proposed method on the Cityscapes dataset, evaluating over different metrics at extremely low bitrates to show that our method outperforms the other state-of-the-art. In particular, at a bitrate of 0.07, CompressNet achieves 22% lower Perceptual Loss and 55% lower Frechet Inception Distance (FID) compared to the deep learning SOTA methods.

eess.IV↗

arXiv2016

Selfie Detection by Synergy-Constraint Based Convolutional Neural Network

Categorisation of huge amount of data on the multimedia platform is a crucial task. In this work, we propose a novel approach to address the subtle problem of selfie detection for image database segregation on the web, given rapid rise in number of selfies clicked. A Convolutional Neural Network (CNN) is modeled to learn a synergy feature in the common subspace of head and shoulder orientation, derived from Local Binary Pattern (LBP) and Histogram of Oriented Gradients (HOG) features respectively. This synergy was captured by projecting the aforementioned features using Canonical Correlation Analysis (CCA). We show that the resulting network's convolutional activations in the neighbourhood of spatial keypoints captured by SIFT are discriminative for selfie-detection. In general, proposed approach aids in capturing intricacies present in the image data and has the potential for usage in other subtle image analysis scenarios apart from just selfie detection. We investigate and analyse the performance of popular CNN architectures (GoogleNet, AlexNet), used for other image classification tasks, when subjected to the task of detecting the selfies on the multimedia platform. The results of the proposed approach are compared with these popular architectures on a dataset of ninety thousand images comprising of roughly equal number of selfies and non-selfies. Experimental results on this dataset shows the effectiveness of the proposed approach.

cs.CV↗