Medical imaging datasets vary widely in modality, dimensionality, and clinical tasks. This diversity typically necessitates single-purpose deep learning models for each domain. We propose Medformer, a unified foundation model for multitask and multimodal medical imaging based on transformer architectures. Medformer uses two specialized modules called Adaptformers—one for input adaptation and one for output adaptation—along with learnable latent embeddings that encode dimensionality (2D vs. 3D), imaging modality (CT, X-ray, microscopy), anatomical region, and task requirements (classification, ordinal regression, etc.). These embeddings guide a shared transformer backbone, enabling broad parameter sharing across heterogeneous tasks. Experiments on the MedMNIST collection, comprising 18 diverse 2D/3D datasets, demonstrate that Medformer can match specialized baselines, particularly for data-scarce tasks, via both multi-task training and self-supervised pretraining. Results suggest that Medformer can serve as a flexible foundation model to unify disparate medical imaging domains within a single architecture.

In this paper we introduce Efficient Dynamic Batch Adaptation (EDBA), which improves on a previous method that works by adjusting the composition and the size of the current batch. Our improvements allow for Dynamic Batch Adaptation to feasibly scale up for bigger models and datasets, drastically improving model convergence and generalization. We show how the method is still able to perform especially well in data-scarce scenarios, managing to obtain a test accuracy on 100 samples of CIFAR-10 of 90.68%, while the baseline only reaches 23.79%. On the full CIFAR-10 dataset, EDBA reaches convergence in ∼120 epochs while the baseline requires ∼300 epochs.

In this paper we introduce Backforward Propagation, a method of completely eliminating Internal Covariate Shift (ICS). Unlike previous methods, which only indirectly reduce the impact of ICS while introducing other biases, we are able to have a surgical view at the effects ICS has on training neural networks. Our experiments show that ICS has a weight regularizing effect on models, and completely removing it enables for faster convergence of the neural network.

This paper presents a novel approach for detecting and finding similar clothing articles in images by combining structural shape detection and texture comparison. We utilized the Mask R-CNN model for structural shape detection and explored color hashing and Local Binary Pattern (LBP) histogram for texture comparison. Experiments were conducted using the DeepFashion2 dataset, demonstrating varying performance across different clothing articles and texture comparison algorithms. While the proposed solution achieved good detection rates for certain clothing articles, it faced challenges in detecting others due to the unequal distribution of articles in the dataset and limitations in texture comparison.

In the recent period, neural networks are used in more and more applications and services across a broader and broader spectrum of industries and domains. One of the applications of neural networks is in artistic content generation, ranging from schematics and improving 3D rendering to AI-powered up-scaling of video-games through DLSS, and image and music generation. In this project, we tried various music generation methods in order to see their limitations, unrefined results, and the most popular generation of audio files.

Millions of people use Twitter every month, which makes it one of the most popular social networks worldwide. Currently, there is an enormous scope market with the potential to be optimized to increase Twitter posts’ popularity and engagement. In this paper, we present a method of predicting the number of likes a given post will receive. We introduce a deep learning model and training procedure that uses both computer vision and natural language processing to reach high accuracy when shown new data. Considered use-cases will show the situations in which our system behaves well and the situations in which we do not yet have a solution to improve the current results.

Electric vehicles present us with a unique set of constraints and considerations when creating routing algorithms. In this paper, we present one such real-world routing algorithm and the specific issues that need to be taken into account when designing such a system. We integrated this algorithm into an Android mobile application named ElectroWay. The code is available on GitHub https://github.com/QuothR/ElectroWay.

As a consequence of the day-to-day increase in energy requirements worldwide, many countries around the world are focusing on implementing renewable energy sources (RES) to become energy independent. Although this concern is usually addressed at a national level, we believe that it also has to be addressed at the individual level, i.e. the consumer segment has to contribute to the shift to a sustainable energy future. This paper aims to provide a tool that helps consumers evaluate potential investment opportunities in renewable energy solutions. The tool focuses on two forms of RES, namely solar energy and wind energy. Several comparison tests were performed on regions from Romania. However, the tool can be also extended to other regions around the world.

Cerebrovascular Accident (CVA) is the second leading cause of death in the world while also being the plurality cause of disability in adults. A definitive factor for survivability and successful recovery of a patient is the time passage from the onset of symptoms to the administration of medical treatment. This paper introduces Stroke Help, a mobile application utilizing various mobile technologies together with Artificial Intelligence algorithms in order to quickly detect CVA in either the user or someone the user is concerned about. The application implements the well known F.A.S.T. test making use of real-time face detection, speech recognition, and other artificial intelligence techniques applied over common sensors found in modern mobile phones. In addition of detecting whether there is a high probability a patient is suffering from a stroke, the application will calculate an approximated Japan Urgent Stroke Triage (JUST) score utilized in identifying the specific type of stroke, very important information for medical staff to potentially reduce the time required to evaluate the patient before beginning the appropriate treatment. We will also present additional crucial functionalities such as notifying contacts, identifying the closest clinics capable of treating CVA, making our solution a complete approach.