Dubravko Ćulibrk

Nikola Jovišić, Milica Škipina, Nicola Dall'Asen et al.

Modern foundation models provide highly expressive visual representations, yet adapting them to high-resolution medical imaging remains challenging due to limited annotations and weak supervision. Mammography, in particular, is characterized by large images, variable multi-view studies and predominantly breast-level labels, making end-to-end fine-tuning computationally expensive and often impractical. We propose Multiple Instance Learning on Precomputed Features (MIL-PF), a scalable framework that combines frozen foundation encoders with a lightweight MIL head for mammography classification. By precomputing the semantic representations and training only a small task-specific aggregation module (40k parameters), the method enables efficient experimentation and adaptation without retraining large backbones. The architecture explicitly models the global tissue context and the sparse local lesion signals through attention-based aggregation. MIL-PF achieves state-of-the-art classification performance at clinical scale while substantially reducing training complexity. We release the code for full reproducibility.

Milica Škipina, Nikola Jovišić, Nicola Dall'Asen et al.

Mammography is the gold standard for the detection and diagnosis of breast cancer. This procedure can be significantly enhanced with Artificial Intelligence (AI)-based software, which assists radiologists in identifying abnormalities. However, training AI systems requires large and diverse datasets, which are often difficult to obtain due to privacy and ethical constraints. To address this issue, the paper introduces MAMmography ensemBle mOdel (MAMBO), a novel patch-based diffusion approach designed to generate full-resolution mammograms. Diffusion models have shown breakthrough results in realistic image generation, yet few studies have focused on mammograms, and none have successfully generated high-resolution outputs required to capture fine-grained features of small lesions. To achieve this, MAMBO integrates separate diffusion models to capture both local and global (image-level) contexts. The contextual information is then fed into the final model, significantly aiding the noise removal process. This design enables MAMBO to generate highly realistic mammograms of up to 3840x3840 pixels. Importantly, this approach can be used to enhance the training of classification models and extended to anomaly segmentation. Experiments, both numerical and radiologist validation, assess MAMBO's capabilities in image generation, super-resolution, and anomaly segmentation, highlighting its potential to enhance mammography analysis for more accurate diagnoses and earlier lesion detection. The source code used in this study is publicly available at: https://github.com/iai-rs/mambo.

Ljubisa Bojic, Matteo Cinelli, Dubravko Culibrk et al.

This paper explores the potential of a multidisciplinary approach to testing and aligning artificial intelligence (AI), specifically focusing on large language models (LLMs). Due to the rapid development and wide application of LLMs, challenges such as ethical alignment, controllability, and predictability of these models emerged as global risks. This study investigates an innovative simulation-based multi-agent system within a virtual reality framework that replicates the real-world environment. The framework is populated by automated 'digital citizens,' simulating complex social structures and interactions to examine and optimize AI. Application of various theories from the fields of sociology, social psychology, computer science, physics, biology, and economics demonstrates the possibility of a more human-aligned and socially responsible AI. The purpose of such a digital environment is to provide a dynamic platform where advanced AI agents can interact and make independent decisions, thereby mimicking realistic scenarios. The actors in this digital city, operated by the LLMs, serve as the primary agents, exhibiting high degrees of autonomy. While this approach shows immense potential, there are notable challenges and limitations, most significantly the unpredictable nature of real-world social dynamics. This research endeavors to contribute to the development and refinement of AI, emphasizing the integration of social, ethical, and theoretical dimensions for future research.

Djordje Batic, Dubravko Culibrk

We present the results of an initial study focused on developing a visual AI solution able to recognize individual dogs in unconstrained (wild) images occurring on social media. The work described here is part of joint project done with Pet2Net, a social network focused on pets and their owners. In order to detect and recognize individual dogs we combine transfer learning and object detection approaches on Inception v3 and SSD Inception v2 architectures respectively and evaluate the proposed pipeline using a new data set containing real data that the users uploaded to Pet2Net platform. We show that it can achieve 94.59% accuracy in identifying individual dogs. Our approach has been designed with simplicity in mind and the goal of easy deployment on all the images uploaded to Pet2Net platform. A purely visual approach to identifying dogs in images, will enhance Pet2Net features aimed at finding lost dogs, as well as form the basis of future work focused on identifying social relationships between dogs, which cannot be inferred from other data collected by the platform.

Enver Sangineto, Moin Nabi, Dubravko Culibrk et al.

In a weakly-supervised scenario object detectors need to be trained using image-level annotation alone. Since bounding-box-level ground truth is not available, most of the solutions proposed so far are based on an iterative, Multiple Instance Learning framework in which the current classifier is used to select the highest-confidence boxes in each image, which are treated as pseudo-ground truth in the next training iteration. However, the errors of an immature classifier can make the process drift, usually introducing many of false positives in the training dataset. To alleviate this problem, we propose in this paper a training protocol based on the self-paced learning paradigm. The main idea is to iteratively select a subset of images and boxes that are the most reliable, and use them for training. While in the past few years similar strategies have been adopted for SVMs and other classifiers, we are the first showing that a self-paced approach can be used with deep-network-based classifiers in an end-to-end training pipeline. The method we propose is built on the fully-supervised Fast-RCNN architecture and can be applied to similar architectures which represent the input image as a bag of boxes. We show state-of-the-art results on Pascal VOC 2007, Pascal VOC 2010 and ILSVRC 2013. On ILSVRC 2013 our results based on a low-capacity AlexNet network outperform even those weakly-supervised approaches which are based on much higher-capacity networks.