This repository contains the code for the Monocular Depth Estimation, a deep learning Challenge on a Unreal Engine Dataset.
The goal of this challenge is to develop a deep neural network to estimate depth information from a single RGB image. The dataset provided includes images of a virtual world generated using Unreal Engine with their corresponding depth maps, the training set has 3469 samples while the validation set has 600 samples.
The proposed model to solve the task follows the commonly used Encoder - Decoder pattern and in particular uses a U-Net style architecture.
All the details about the model created and the results obtained are included in the final report [Deep_Learning_Challenge_Final_Report.pdf].