Presentation

During my second year at EPITA, the Fooo team has undertaken an ambitious project: a software for Optical Music Recognition (OMR, aka Music OCR).
The project was written mainly in C and OCaml.

Contributions

I was in charge of developing a neural network for character recognition.

As a first attempt I developed a classical multilayer perceptron (MLP).
However the results obtained on the MNIST database of handwritten digits were not good enough. The first major problem of a MLP is its low resilience to shifting, scaling or other forms of distortion. The second major problem is the fact that the geometry of the input image is not taken into account. In order to solve this problem, the input of the MLP is not the raw image but rather the output of a features extractor.

After doing some research on other classification methods and image characterization techniques I decided to implement a simple but efficient variant of MLPs: Convolutional Neural Networks (CNN) introduced by Yann LeCun and Yoshua Bengio in this paper.
The image is divided in several receptive fields, for instance overlapping regions of 5x5 pixels. The second layer is obtained by applying several convolutions on the input image. The values of the convolution kernel are computed during the learning phase of the neural network. The application of a convolution on the input image forms a features map in the subsequent layer. Generally, 4 or 6 features maps are used in the first hidden layer.
The second hidden layer is a subsampling layer, a local average is computed for each value of the previous feature map. Using this technique, the classifier achieves some form of translation invariance.

This convolution/subsampling alternation is used twice before the output layer.
We therefore have this order: input-convolution-subsampling-convolution-subsampling-output.

Using Convolutional Neural Networks we achieved good classification results on the MNIST database. After preprocessing of the input data (for example distortion) we achieved very goods results: around 97% on the test set.

The final report for the project can be found here (in French).
A good explanation on CNNs can be found here.

Presentation

Fooo was my very first computer science project. It was developed in 8 months during my first year at EPITA, a french graduate school of computer science and advanced technologies and succeeded in finishing top project of the year.
In a team of 4 students (with Christopher Chedeau, Vladimir Nachbaur and Alban Perillat-Merceroz) we developed a Real Time Strategy (RTS) game largely inspired from Warcraft III. However, textures, models and icons were not taken from MDX/MDL/BLP files of this game. Instead we used non-copyrighted fan-made art, with the approval of the authors.

The core of the game was developed in Delphi, the 3D engine was developed with OpenGL and the interface is in Lua.
Source files, reports (in French) and screenshots can be found on our website.

Video

We have made a video presenting all the features of our project, check it out !

Contribution

As mentioned, Fooo was my first large project experience. The project quickly started after being introduced to programming through the functional language OCaml and later Delphi (language inspired by Object Pascal).

Developing a video game from scratch during the first year seems to be rather a daunting project. The policy of our school to help us develop our project was rather simple: there was no such policy. Of course we had general algorithms courses and smaller Delphi project in the meantime but concerning project design, pathfinding, 3D engine, networking... we had to figure most of it by ourselves by reading books, forums, tutorials...

I was mostly in charge of the core of the game: movement, orders (e.g. patrol, attack), spells, resources generation and usage, unit creation, unit death, building creation, etc.
I also contributed to the interface in order to associate each user action (like mouse selection, button click, etc) to an action in the game's engine.