LGF Seminar – Abbass ZeinEddine – May 16, 2024

Title

Enhancing Object Detection in Distorted Environments: A Seamless Integration of Classical Image Processing with Deep Learning Models

Abstract

In this presentation, we unveil a novel method to enhance object detection in distorted images. Our approach integrates histogram equalization and other classical image processing techniques seamlessly, while leveraging depth-wise convolution and deformable convolution to mitigate distortion effects. Through multi experiments, we demonstrate significant improvements over conventional RGB-only approaches, showcasing our contributions towards more accurate object detection in
challenging real-world scenarios.