We present SITUATE, a novel dataset designed for training and evaluating Vision Language Models on counting tasks with spatial constraints. The dataset bridges the gap between simple 2D datasets like VLMCountBench and often ambiguous real-life datasets like TallyQA, which lack control over occlusions and spatial composition. Experiments show that our dataset helps to improve generalization for out-of-distribution images, since a finetune of Qwen VL 2.5 7B on SITUATE improves accuracy on the Pixmo count test data, but not vice versa. We cross validate this by comparing the model performance across established other counting benchmarks and against an equally sized fine-tuning set derived from Pixmo count.

There is a huge demand for trying on clothing at home. Recent methods to capture your figure mostly work in a 2d image plane and despite recent improvements of available technology the simulation of clothing is still not satisfactory. Especially the rendering of different clothing sizes is still a major challenge, which is only addressed by COTTON [1]. We propose an improvement to this approach by adding more control over the image generation process. For this we employ a special type of conditional diffusion model, namely ControlNet, and take keypoints of the fashion as conditional input. (This work is a part of the project (M4-SKI) has been supported and funded by the European Regional Development Fund (ERDF)).

In this paper we present a stable-diffusion based zero-shot approach to realistically transform the image of a

human body into a more fit version of that depicted person. Therefore we combine a modified stable diffusion

model with inpainting techniques and incorporated constraints. We introduce a prototype which allows users to

upload a photo and visualize a more fit version of themselves. We evaluated our approach in various experiments

and focused on the applicability and effectiveness of these techniques, with attention to gender-specific results.

This work contributes to the fields of computer vision and generative AI by demonstrating practical applications

and identifying areas for improvement in realistic body transformation visualizations. (This work is a part of the project (M4-SKI) has been supported and funded by the European Regional Development Fund (ERDF)).

The simulation of clothing for a virtual try on is still a challenging task, especially if the customer wants to use state of the art technology. To address this, we employ a 2D plane to process customer images. Specifically, we utilize a neural network, notably an autoencoder, to render so called fashion landmarks. As input we use human keypoints that represent the model poses and our fashion landmarks of the clothing from stock photos to generate fashion landmarks in the desired pose. These can be utilized by additional algorithms to adapt the clothing length or width.

This project has been funded by the European Regional Development Fund (EFRE).

EEG classification is a promising approach to facilitate the life of handicapped people and to generate future human-computer-interfaces. In this paper we want to compare the effectiveness of current state of the art deep learning techniques for EEG classification. Therefore, we applied different approaches on various datasets and did a crosscomparison of the results in order to get more knowledge on the generalization capabilities. Additionally, we created a new EEG dataset and made it available for further research.