arXiv subjects

Dabbrata Das

Publications and source records attributed to Dabbrata Das.

4 recordsLinked to original sources

arXiv2025

Edge-Enhanced Vision Transformer Framework for Accurate AI-Generated Image Detection

The rapid advancement of generative models has led to a growing prevalence of highly realistic AI-generated images, posing significant challenges for digital forensics and content authentication. Conventional detection methods mainly rely on deep learning models that extract global features, which often overlook subtle structural inconsistencies and demand substantial computational resources. To address these limitations, we propose a hybrid detection framework that combines a fine-tuned Vision Transformer (ViT) with a novel edge-based image processing module. The edge-based module computes variance from edge-difference maps generated before and after smoothing, exploiting the observation that AI-generated images typically exhibit smoother textures, weaker edges, and reduced noise compared to real images. When applied as a post-processing step on ViT predictions, this module enhances sensitivity to fine-grained structural cues while maintaining computational efficiency. Extensive experiments on the CIFAKE, Artistic, and Custom Curated datasets demonstrate that the proposed framework achieves superior detection performance across all benchmarks, attaining 97.75% accuracy and a 97.77% F1-score on CIFAKE, surpassing widely adopted state-of-the-art models. These results establish the proposed method as a lightweight, interpretable, and effective solution for both still images and video frames, making it highly suitable for real-world applications in automated content verification and digital forensics.

cs.CV↗

arXiv2025

A Dual-Layer Image Encryption Framework Using Chaotic AES with Dynamic S-Boxes and Steganographic QR Codes

This paper presents a robust image encryption and key distribution framework that integrates an enhanced AES-128 algorithm with chaos theory and advanced steganographic techniques for dual-layer security. The encryption engine features a dynamic ShiftRows operation controlled by a logistic map, variable S-boxes generated from a two-dimensional Henon map for substitution and key expansion, and feedback chaining with post-encryption XOR diffusion to improve confusion, diffusion, and key sensitivity. To address secure key delivery, the scheme introduces dual-key distribution via steganographically modified QR codes. A static key and an AES-encrypted dynamic session key are embedded with a covert hint message using least significant bit (LSB) steganography. This design ensures the dynamic key can only be decrypted after reconstructing the static key from the hidden message, offering multi-factor protection against interception. Experimental results demonstrate the framework outperforms existing chaos-based and hybrid AES methods, achieving near-ideal entropy (7.997), minimal pixel correlation, and strong differential resistance with NPCR (>99.6%) and UACI (50.1%). Encrypted images show uniform histograms and robustness against noise and data loss. The framework offers a scalable, secure solution for sensitive image transmission in applications such as surveillance, medical imaging, and digital forensics, bridging the gap between cryptographic strength and safe key distribution.

cs.CR↗

arXiv2025

Real-Time Assistive Navigation for the Visually Impaired: A Scalable Approach for Indoor and Outdoor Mobility

Navigating unfamiliar environments remains one of the most persistent and critical challenges for people who are blind or have limited vision (BLV). Existing assistive tools often rely on online services or APIs, making them costly, internet-dependent, and less reliable in real-time use. To address these limitations, we propose PathFinder, a novel mapless mobile phone-based navigation system that operates fully offline. Our method processes monocular depth images and applies an efficient pathfinding algorithm to identify the longest, clearest obstacle-free route, ensuring optimal navigation with low computational cost. Comparative evaluations show that PathFinder reduces mean absolute error (MAE), speeds decision-making, and achieves real-time responsiveness indoors and outdoors. A usability study with 15 BLV participants confirmed its practicality, where 73% learned to operate it in under a minute, and 80% praised its accuracy, responsiveness, and convenience. Despite challenges in complex indoor layouts and low light, PathFinder offers a low-cost, scalable, reliable alternative.

cs.HC↗

arXiv2024

Enhanced Encoder-Decoder Architecture for Accurate Monocular Depth Estimation

Estimating depth from a single 2D image is a challenging task due to the lack of stereo or multi-view data, which are typically required for depth perception. In state-of-the-art architectures, the main challenge is to efficiently capture complex objects and fine-grained details, which are often difficult to predict. This paper introduces a novel deep learning-based approach using an enhanced encoder-decoder architecture, where the Inception-ResNet-v2 model serves as the encoder. This is the first instance of utilizing Inception-ResNet-v2 as an encoder for monocular depth estimation, demonstrating improved performance over previous models. It incorporates multi-scale feature extraction to enhance depth prediction accuracy across various object sizes and distances. We propose a composite loss function comprising depth loss, gradient edge loss, and Structural Similarity Index Measure (SSIM) loss, with fine-tuned weights to optimize the weighted sum, ensuring a balance across different aspects of depth estimation. Experimental results on the KITTI dataset show that our model achieves a significantly faster inference time of 0.019 seconds, outperforming vision transformers in efficiency while maintaining good accuracy. On the NYU Depth V2 dataset, the model establishes state-of-the-art performance, with an Absolute Relative Error (ARE) of 0.064, a Root Mean Square Error (RMSE) of 0.228, and an accuracy of 89.3% for $\delta$ < 1.25. These metrics demonstrate that our model can accurately and efficiently predict depth even in challenging scenarios, providing a practical solution for real-time applications.

cs.CV↗