Martin Thoma

Martins Python Utilities (mpu) is a collection of utility functions and classes with no other dependencies.

The total size of the package will never be bigger than 10 MB and currently it is 15.7 kB in zipped form. This makes it a candidate to include into AWS Lambda projects.

One of the simplest sensors which can be used in self-driving cars are cameras which give a 2D RGB video stream. Having the position, camera characteristics and other sensors the information which class each pixel belongs to gives valuable clues about the environment. Typical classes are street, sky, car, pedestrian, traffic sign, bicyclist, ...

In this project, we developed a convolutional neural network which is capable of segmenting street in real time. We used Python and the Lasagne… Show more

One of the simplest sensors which can be used in self-driving cars are cameras which give a 2D RGB video stream. Having the position, camera characteristics and other sensors the information which class each pixel belongs to gives valuable clues about the environment. Typical classes are street, sky, car, pedestrian, traffic sign, bicyclist, ...

In this project, we developed a convolutional neural network which is capable of segmenting street in real time. We used Python and the Lasagne framework, which is based on Theano. Show less

A project to automatically recognize handwritten mathematical symbols. It makes use of multilayer Perceptrons (feed-forward neural networks) to classify the input into 369 classes.

Technologically, the project

Sorting is an essential in many tasks: That could be ore in rocks, separating spoiled from good food (e.g. meat or mushrooms), trash and similar. This can be done quickly if one can identify the pieces that need to be separated. Quick and reliable identification, however, is not that easy. To do so, one can sometimes put the material under light of specific frequencies under which the differences become obvious. Finding the best frequency is a research task which needs to be done once. This… Show more

Sorting is an essential in many tasks: That could be ore in rocks, separating spoiled from good food (e.g. meat or mushrooms), trash and similar. This can be done quickly if one can identify the pieces that need to be separated. Quick and reliable identification, however, is not that easy. To do so, one can sometimes put the material under light of specific frequencies under which the differences become obvious. Finding the best frequency is a research task which needs to be done once. This project supports scientist at Fraunhofer IOSB to run the experiments and store the data.

On the technological side, C# / ASP.NET was used. Show less