Semantic segmentation with Deep Learning

For quantitative analysis, cells in microscopic images need to be segmented from the image. In this project it model the problem in terms of semantic segmentation and approach it by means of deep learning, and more specifically with the U-Net architecture.

Auto-calibration using LIDAR and Visual SLAM

In this project it is proposed a method for the calibration of a vehicle equipped with a 3DLIDAR and cameras. Combining known vehicle trajectory with the poses from camera and LIDAR, the idea is to propose a non-linear minimization framework to solve the relative transformation between the LIDAR sensor and the camera.

Automatic music transcription

This project aims to automatically transcripts music notes based on a scoring metric. The signal is divided into successive time slots of small duration, on which spectral analysis is performed by Fourier transform.

Coin Counter

This project presents a coin counter system based on Convolutional Neural Networks (CNNs) and Digital Image Processing techniques. The framework's uses images taken by a cellphone camera as input and output the quantity of coins and the amount of money. Note: it should be noted that the system has been tests ONLY on Colombian coins.