Hey, I'm Sebastian. I love building programs that solve real life problems (sometimes :-)
I began my programming journey with a quirky quest: building my own random number generator.
Since then, i regularly discover new stuff to dive in/break my head on for some weeks.
I was born in Bordeaux and I enjoy traveling.
Feel free to email me - sebastian.s.straut@protonmail.com
This project is a simple API to get the tram stop timetables from the TBM API.
In this case, the tram stop is Doyen Brus towards Bordeaux.
- Add the possibility to choose the stop
- Add cache system to avoid calling the API too often
- Add system to create a widget to display the departures
This project uses the yt-dlp library to download Youtube videos,
a ajax onsubmit button loads a php file in the background which sends
the url to the python script via ssh command
then javascript to display the video in the browser, or download it.
- Add the possibility to download multiple videos at once
- Add the possibility to download videos from other web sites
This project, in collaboration with Eva Morton and Yves Appriou, is part of my final year school project.
This program applies Gaussian random noise to a given image, then uses Markov chains to reconstruct the original image.
In fact, this program's goal is to minimize the overall energy of the image. This energy is calculated by choosing a random pixel in the image,
reversing this pixel's value (in the black and white case), and then calculating the sum of the neighboring pixels' values.
Metropolis or Gibbs sampling is applied to that sum, which gives us a two-dimensional list containing the probability of accepting the new pixel value.
This program's iteration is stopped when the energy is stable or when the number of iterations is reached.
Take a look at the paper
Metropolis sampling visualization
This program takes a .wav file as input and returns the numbers typed in. It takes the audio file, and applies a function to detect the intervals of silence.
Then it uses the FFT to detect the frequencies of the numbers typed in, One low frequency and one high frequency.
With the findpeaks library, we detect the peaks in the frequencies and displays them.
And by using the DTMF matrix, we decode the numbers typed in with an error rate of 10Hz.
- Enhence the parsite noice reduction
- Transport this program to keyboard typing detection
BTF are really heavy data, made out of multiple photos from different camera/light angles.
Materials apparence in many cases, have redundant informations, thus, the idea behind this Neural Network,
is the compression of this redundant data in a multiple dimensions latent space, so we can recreate all
points of view from a material with only our nn and a light/camera direction.
Project conducted during my intership at Centre Inria de l'Université de Bordeaux, under the supervision of Patrick Reuter
This project uses computer vision to detect and track hand movements in real-time. Using MediaPipe's hand tracking library, it can identify hand landmarks and gestures. Optimized by multithreadding, this program can recognize different hand positions and movements, making it useful for gesture-based control systems.
- Real-time hand detection and tracking
- Gesture recognition
- Multiple hand support
