Projects & Publications

Simulating Depth Measuring Sensors for Autonomous Learning and Benchmarking, Master thesis, 2018.

The master thesis was exploring usage of CycleGAN in the context of depth sensors for autonomous driving, especially Velodyne LiDAR. Our main goal was to be able to generate LiDAR data of real-world quality by the means of first simulating them via game engine and then tweaking them by CycleGANs in order to correspond to real-world data as closely as possible.


M. Krčál, O. Švec, M. Bálek, O. Jašek: Deep Convolutional Malware Classifiers Can Learn from Raw Executables and Labels Only, ICLR 2018

We published this paper while I was with Avast as a research intern. The paper proposes novel architecture of neural network for classifying malware in an end-to-end fashion while operating only on executables and makes heavy use of enormous dataset of malicious software that Avast operates with.

My contribution to the paper were rather technical as I was mostly doing programming of all the required tools and datasets.


Detecting Objects for Autonomous System Verification, Bachelor thesis, 2016

The bachelor thesis was about exploring available options for detecting objects using convolutional networks and main focus was on networks of Faster-RCNN architecture which were rather novel at the time.