Computational Depth-of-Field

This project focuses on the implementation of Computational Depth-of-Field using various techniques and tools. The primary objective is to simulate the depth-of-field effect, which is a characteristic of optical systems, using computational methods.

Depth-of-Field (DoF) refers to the range in a scene where the objects appear acceptably sharp in an image. In photography, it's used to draw attention to specific subjects in the frame by blurring the foreground and/or background. This project aims to achieve this effect computationally, allowing for post-processing adjustments to the perceived depth in an image.

Files

• bilateral_filter.py: A module containing functionality related to the bilateral filtering process. This filtering method is used to reduce noise while preserving edges.
• depth_map_CNN.py: A module related to estimating the depth map using the MiDaS 3.1 dpt_swin2_tiny_256 model CNN.
• depth_map.py: A module for generating and working with depth maps by solving the constrained Poisson equation.
• helpers.py: A utility module that contains helper functions, which are commonly used across different parts of the application.
• ken_burns.py: A module related to the Ken Burns effect.
• main.py: The main entry point of the application.
• ui.py: A module related to the user interface of the application.

Setup

This code is written in Python 3.10.8 and requires the packages listed in requirements-cpu.yaml (for CPU-based execution) or requirements-gpu.yaml (for GPU usage).

CPU-based execution

# Create a virtual environment using conda
conda env create --file environment-cpu.yaml

# Activate the virtual environment
conda activate cpt-dof-cpu

GPU usage

# Create a virtual environment using conda
conda env create --file environment-gpu.yaml

# Activate the virtual environment
conda activate cpt-dof-gpu

Usage

python src/main.py

Use the associated command-line arguments to specify parameters:

• Initial image file path to load: -F/--file
• Enable model optimization for CUDA when using the CNN depth map estimation (only works if CUDA exists): -o/--optimize
• Warping factor (in %) for the Ken Burns effect: -w/--warping-factor
• Near plane distance (millimeters) for the Ken Burns effect: -n/--near-plane
• Far plane distance (millimeters) for the Ken Burns effect: -f/--far-plane
• Duration of the video (seconds) for the Ken Burns effect: -d/--duration-video
• Enable anaglyph mode when using the Ken Burns effect: -a/--anaglyph

Note that all the possible command line arguments can be consulted using python src/main.py -h or python src/main.py --help.

Examples

In this section can be found examples outputed by this application. For more examples and detailed view of this project there is also a report and a video.

Example 1 - Arch

Legend: 3D Ken-Burns effect applied on arch image.

Example 2 - Firefighter

Legend: 3D Ken-Burns effect applied on firefigther image.

Example 3 - Miami Rescue

Legend: 3D Ken-Burns effect applied on Miami rescue image.

Example 4.1 - Art

Legend: 3D Ken-Burns effect applied on art image.

Example 4.2 - Art

Legend: Stereo Ken-Burns effect applied on art image produced by using python src/main.py -a. Put your 3D glasses on!