Publications (reverse chronological order)
3. Adversarial Augmentation for Enhancing Classification of Mammography Images (preprint)
Lukas Jendele*, Ondrej Skopek*, Anton S. Becker, Ender Konukoglu. “Adversarial Augmentation for Enhancing Classification of Mammography Images”, 2019.
* equal contributions
On the example of a binary image classification task (breast cancer recognition), we show that pretraining a generative model for meaningful image augmentation helps enhance the performance of the resulting classifier. By augmenting the data, performance on downstream classification tasks could be improved even with a relatively small training set. We show that this “adversarial augmentation” yields promising results compared to classical image augmentation on the example of breast cancer classification.
2. Injecting and removing malignant features in mammography with CycleGAN: Investigation of an automated adversarial attack using neural networks (preprint)
Anton S. Becker, Lukas Jendele*, Ondrej Skopek*, Nicole Berger, Soleen Ghafoor, Magda Marcon, Ender Konukoglu. “Injecting and removing malignant features in mammography with CycleGAN: Investigation of an automated adversarial attack using neural networks”, 2018.
* equal contributions
We train a cycle-consistent generative adversarial network (CycleGAN) on mammographic data to inject or remove features of malignancy, and determine whether these AI-mediated attacks can be detected by radiologists.
Ondrej Skopek, Roman Bartak. “TransportEditor – Creating and Visualising Transportation Problems and Plans.” Prague, 2017.
Video: Demo of an older version of TransportEditor.
Slides accompanying the presentation.
Master thesis: To Be Announced
Bachelor thesis: Planning for Transportation Problems
“Planning for Transportation Problems”, thesis for my bachelor programme at MFF UK. Supervised by Prof. RNDr. Roman Barták, Ph.D.. Source code and supporting documents can be found on GitHub and GitLab. Poster. Official university websites.
Reports (reverse chronological order)
6. Story Cloze (Course: Natural Language Understanding)
Choosing the correct short story ending out of two possibilities (Story Cloze Task), where the training data only contains the correct endings. Achieved close to state of the art results at the time. Course project for the Natural Language Understanding course at ETH Zurich. Co-authored with Lukas Jendele, Vasily Vitchevsky, and Michael Wiegner. Paper. GitHub repository.
5. Tweet Sentiment Analysis (Course: Computational Intelligence Lab)
Sentiment analysis of a large dataset of tweets using weakly supervised learning. The dataset was scraped and labeled automatically based on the presence of positive or negative emoji. We perform an extensive study of different models and provide a comparison. Course project for the Computational Intelligence Lab course at ETH Zurich. Co-authored with Lukas Jendele, Larissa Laich, and Michael Wiegner. Paper. GitHub repository.
4. Eye Gaze Estimation (Course: Machine Perception)
3. Stupid, Crazy, Love. (Course: Network Security)
2. Data structures for high-dimensional search (Course: Algorithms & Data Structures)
High school scientific paper (in Slovak) about “Recognition of cars and calculation of speed using computer vision and the OpenCV library” (“Rozpoznávanie áut a výpočet rýchlosti jazdy pomocou počítačového videnia a knižnice OpenCV”), which won 4th place at the national round of the High School Scientific Activity contest in Nové Zámky in category 11 - Computer Science (Stredoškolská odborná činnosť).