Table of Contents
Main research interests
Machine Learning in Robotics
With Prof. Luca Gambardella, I co-lead the IDSIA Robotics Lab.
Applied Deep Learning
I am leading several applied funded research projects on machine learning applications to industrial tasks (1M+ CHF funding).
I am also collaborating with colleagues, applying Machine Learning to various disciplines such as Optical Networks (with Cristina Rottondi)
Biomedical image analysis and visualization
My past activities in the field of Biomedical image analysis, featuring state-of-the-art results on many datasets including membrane segmentation in EM images, mitosis detection in histopathology images, retinal blood vessel segementation, segmentation of embryos and zygotes in hoffmann modulation contrast microscopy images and stacks.
I also developed a cool (and useful) visualization technique for visual markers on Z-stacks.
Other cool stuff (past)
- Analysis and re-animation of motion-blurred objects was my main research topic during my PhD. An interesting problem which involves elegant, novel theories, and provides counterintuitive results. There are also various practical applications.
- Speedlines: Drawing Motion without Understanding It, i.e. synthesizing comicbook-like speedlines on videos, with only low-level processing.
- 3D Reconstruction of Canal Surfaces from a single perspective image.
- Convolutional Deep Neural Networks are surprisingly good for image segmentation and object detection tasks; but applying these networks to every patch of a large image is computationally expensive: here's our fast scanning algorithm which speeds this up by several orders of magnitude (a precursor to fully convolutional networks).