VISION

This repository contains practical works (PW) I completed, along with slides from a presentation I co-delivered with another student, as part of my master's course in computer vision (https://perso.ensta-paris.fr/~manzaner/Cours/IMA/VISION/).

Table of Contents

• PW n°1: GC Disparity
• PW n°2: Face Alignment
• PW n°3: Fundamental Matrix
• PW n°4: Object Tracking
• PW n°5: Optical Flow
• PW n°6: Panorama Stitching
• PW n°7: Seeds
• Oral Presentation

PW n°1: GC Disparity

This section contains C++ code and instructions for computing disparity using Graph Cuts.

• Code: GCDisparity.cpp, exampleGC.cpp
• Instructions: PDF

PW n°2: Face Alignment

This section includes Python code (Jupyter Notebook) and instructions for face alignment.

• Code: PW_face_alignment.ipynb
• Instructions: PDF

PW n°3: Fundamental Matrix

This section contains C++ code and instructions for computing the fundamental matrix.

• Code: Fundamental.cpp
• Instructions: PDF

PW n°4: Object Tracking

This section includes Python code and instructions for object tracking using MeanShift.

• Code: code
• Instructions: PDF

PW n°5: Optical Flow

This section contains Python code (Jupyter notebook) and instructions for computing the optical flow of images, and a report presenting the results.

• Instructions: PDF
• Code: code
• Report: PDF

PW n°6: Panorama Stitching

This section contains C++ code and instructions for stitching images to create a panorama.

• Code: Panorama.cpp
• Instructions: PDF

PW n°7: Seeds

This section contains C++ code and instructions related to the Seeds algorithm.

• Code: Seeds.cpp
• Instructions: PDF

Oral Presentation

This presentation, entitled 'Particle Tracking with Multiple Event Cameras', I did with another student (Nils Aurdal), focused on using event-based cameras and the Kalman filter for 3D tracking of particles in a wind tunnel. It highlighted the advantages of event-based cameras, such as their high dynamic range and temporal resolution. The presentation also discussed various filtering methods for data processing and different techniques for reconstructing particle tracks in 3D space.

• Presentation Slides

Usage

Python projects (PW n°2, 4 & 5)

Step 1: Clone the Repository

Clone the VISION repository to your local machine using the following commands:

git clone https://github.com/pictoune/VISION.git
cd VISION

Step 2: Create the required Conda Environment

Set up the required environment using Conda:

conda env create -f environment.yml -n VISION_env

Step 3: Running the code

• If the practical work you are interested in is written in .py files, you must first activate the conda environment:

  conda activate TADI_env

  then you can run it:

  python <script_name>.py

• Otherwise if it is written in a jupyter notebook, you need to initiate the notebook first:

    jupyter-notebook

  Once Jupyter Notebook is open, navigate to the notebook you want to run. Then, change the kernel to the VISION environment: Go to Kernel -> Change kernel -> Python [conda env:VISION_env]. Then you can run the cells.

C++ projects (PW n°1, 3, 6 & 7)

To test the code of the practical work done in C++, you just need to make sure that the Imagine++ library is installed (see https://imagine.enpc.fr/~monasse/Imagine++/).

License

This project is open source and available under the MIT License.

Feel free to explore the projects and reach out if you have any questions or suggestions.