Deep Learning

This article is not finished and will be updated. The research group I work with has access to a small GPU cluster, which occasionally sits idle. To avoid wasting valuable compute resources (IDLE GPUs essentially burn money through opportunity costs), I decided to train a German GPT-2-style model from scratch, using only German text. Existing German models available on Hugging Face have 137M parameters and a context length of 1024 tokens1, which is quite limited compared to recently released …

Our paper Deep learning-based harmonization and super-resolution of Landsat-8 and Sentinel-2 images, which is based on the masters thesis of my colleague Venkatesh Thirugnana Sambandham, has been published in the ISPRS Journal of Photogrammetry and Remote Sensing. This work is an extension of our previous workshop paper on transformers for satellite homogenization. In summary, we find that a simple UNet model provides surprisingly good performance for the satellite homogenization task. We …

Deep Neural Networks are black-boxes: they map some input to some output, and we can make them do this surprisingly well. However, we usually have no idea how this mapping works. Particularly Convolutional Neural Networks (CNNs), which employ “convolutions” as filters, achieved some impressive results (before Vision Transformers came along). Filter Visualization can help us understand what kind of patterns the convolutional filters in CNNs detect. Why would we want to do it? § …

Masters Thesis (PDF). If an AI gives you a weird explanation for its prediction, you should remain septical about the accuracy of the prediction. Sounds reasonable? This was the general idea of my masters thesis, which was originally titled Self-Assessment of Visual Recognition Systems based on Attribution. Today, I would call it Explanation-based Anomaly Detection in Deep Neural Networks. The general idea was to use attribution-based explanation methods to detect anomalies (such as unusual …