Deep Learning Based Monocular Depth Estimation for Object Distance Inference in 2D Images

Abstract

Monocular depth estimation, a process of predicting depth from a single 2D image, has seen significant advancements due to the proliferation of deep learning techniques. This research focuses on leveraging deep learning for monocular depth estimation to infer object distances accurately in 2D images. We explore various convolutional neural network (CNN) architectures and transformer models to analyze their efficacy in predicting depth information. Our approach involves training these models on extensive datasets annotated with depth information, followed by rigorous evaluation using standard metrics. The results demonstrate substantial improvements in depth estimation accuracy, highlighting the potential of deep learning in enhancing computer vision tasks such as autonomous driving, augmented reality, and robotic navigation. This study not only underscores the importance of model architecture but also investigates the impact of training data diversity and augmentation strategies. The findings provide a comprehensive understanding of the current state-of-the-art in monocular depth estimation, paving the way for future innovations in object distance inference from 2D images. By providing a detailed analysis of various models and their performance, this research contributes to a better understanding of monocular depth estimation and its potential for real-world applications, paving the way for future advancements in object distance inference from 2D images.