Ibrahim Delibasoglu

Ibrahim Delibasoglu

The rapid evolution of generative models has enabled the creation of hyper-realistic facial deepfakes, exposing a critical vulnerability in modern digital forensics: the inability of detectors to generalize to unseen manipulation techniques. Traditional networks suffer from representation collapse, overfitting to localized artifact fingerprints of specific training generators. This work investigates whether modern Vision Foundation Models can serve as generalizable, out-of-the-box feature extractors capable of tracking forensic anomalies across entirely unseen generative manifolds. We conduct a systematic cross-domain evaluation comparing three foundational learning paradigms: fully supervised macro-semantic features (RoPE-ViT), pure self-supervised geometric features (DINOv3), and multi-teacher agglomerative representations (NVIDIA C-RADIOv4-H). By deploying frozen backbones subjected to downstream linear probing, we map the performance limitations of these architectures on the challenging DF40 benchmark. Our empirical findings expose the intrinsic trade-offs between pre-training paradigms and parameter scale, proving that while foundation models retain high discriminative capabilities for entire face synthesis, localized face editing techniques expose fundamental boundaries in linear probe evaluation structures. Source code and model weights are available in http://github.com/mribrahim/deepfake

Ibrahim Delibasoglu, Sanjay Chakraborty, Fredrik Heintz

Time series forecasting is an important challenge with significant applications in areas such as weather prediction, stock market analysis, scientific simulations and industrial process analysis. In this work, we introduce LMS-AutoTSF, a novel time series forecasting architecture that incorporates autocorrelation while leveraging dual encoders operating at multiple scales. Unlike models that rely on predefined trend and seasonal components, LMS-AutoTSF employs two separate encoders per scale: one focusing on low-pass filtering to capture trends and the other utilizing high-pass filtering to model seasonal variations. These filters are learnable, allowing the model to dynamically adapt and isolate trend and seasonal components directly in the frequency domain. A key innovation in our approach is the integration of autocorrelation, achieved by computing lagged differences in time steps, which enables the model to capture dependencies across time more effectively. Each encoder processes the input through fully connected layers to handle temporal and channel interactions. By combining frequency-domain filtering, autocorrelation-based temporal modeling, and channel-wise transformations, LMS-AutoTSF not only accurately captures long-term dependencies and fine-grained patterns but also operates more efficiently compared to other state-of-the-art methods. Its lightweight design ensures faster processing while maintaining high precision in forecasting across diverse time horizons. The source code is publicly available at \url{http://github.com/mribrahim/LMS-TSF}

Ibrahim Delibasoglu

Mixture of Experts (MoE) architectures enable efficient scaling of neural networks but suffer from expert collapse, where routing converges to a few dominant experts. This reduces model capacity and causes catastrophic interference during adaptation. We propose the Spectrally-Regularized Mixture of Experts (SR-MoE), which imposes geometric constraints on the routing manifold to enforce structural modularity. Our method uses dual regularization: spectral norm constraints bound routing function Lipschitz continuity, while stable rank penalties preserve high-dimensional feature diversity in expert selection. We evaluate SR-MoE across architectural scales and dataset complexities using modular one-shot adaptation tasks. Results show that traditional linear gating fails with increasing depth (accuracy drops up to 4.72% due to expert entanglement), while SR-MoE maintains structural integrity (mean interference -0.32%). Our spectral constraints facilitate positive knowledge transfer, enabling localized expert updates without global performance decay. SR-MoE provides a general solution for building high-capacity, modular networks capable of stable lifelong learning.

Sanjay Chakraborty, Ibrahim Delibasoglu, Fredrik Heintz

Time series forecasting is a crucial challenge with significant applications in areas such as weather prediction, stock market analysis, and scientific simulations. This paper introduces an embedded decomposed transformer, 'EDformer', for multivariate time series forecasting tasks. Without altering the fundamental elements, we reuse the Transformer architecture and consider the capable functions of its constituent parts in this work. Edformer first decomposes the input multivariate signal into seasonal and trend components. Next, the prominent multivariate seasonal component is reconstructed across the reverse dimensions, followed by applying the attention mechanism and feed-forward network in the encoder stage. In particular, the feed-forward network is used for each variable frame to learn nonlinear representations, while the attention mechanism uses the time points of individual seasonal series embedded within variate frames to capture multivariate correlations. Therefore, the trend signal is added with projection and performs the final forecasting. The EDformer model obtains state-of-the-art predicting results in terms of accuracy and efficiency on complex real-world time series datasets. This paper also addresses model explainability techniques to provide insights into how the model makes its predictions and why specific features or time steps are important, enhancing the interpretability and trustworthiness of the forecasting results.

Sanjay Chakraborty, Ibrahim Delibasoglu, Fredrik Heintz

Large pre-trained models have demonstrated remarkable capabilities across domains, but their effectiveness in time series forecasting remains understudied. This work empirically examines whether pre-trained large-scale time series models (LSTSMs) trained on diverse datasets can outperform traditional non-pretrained small-scale transformers in forecasting tasks. We analyze state-of-the-art (SOTA) pre-trained universal time series models (e.g., Moirai, TimeGPT) alongside conventional transformers, evaluating accuracy, computational efficiency, and interpretability across multiple benchmarks. Our findings reveal the strengths and limitations of pre-trained LSTSMs, providing insights into their suitability for time series tasks compared to task-specific small-scale architectures. The results highlight scenarios where pretraining offers advantages and where simpler models remain competitive.

Ibrahim Delibasoglu, Fredrik Heintz

Explainability in time series forecasting is essential for improving model transparency and supporting informed decision-making. In this work, we present CrossScaleNet, an innovative architecture that combines a patch-based cross-attention mechanism with multi-scale processing to achieve both high performance and enhanced temporal explainability. By embedding attention mechanisms into the training process, our model provides intrinsic explainability for temporal saliency, making its decision-making process more transparent. Traditional post-hoc methods for temporal saliency detection are computationally expensive, particularly when compared to feature importance detection. While ablation techniques may suffice for datasets with fewer features, identifying temporal saliency poses greater challenges due to its complexity. We validate CrossScaleNet on synthetic datasets with known saliency ground truth and on established public benchmarks, demonstrating the robustness of our method in identifying temporal saliency. Experiments on real-world datasets for forecasting task show that our approach consistently outperforms most transformer-based models, offering better explainability without sacrificing predictive accuracy. Our evaluations demonstrate superior performance in both temporal saliency detection and forecasting accuracy. Moreover, we highlight that existing models claiming explainability often fail to maintain strong performance on standard benchmarks. CrossScaleNet addresses this gap, offering a balanced approach that captures temporal saliency effectively while delivering state-of-the-art forecasting performance across datasets of varying complexity.

Ibrahim Delibasoglu

This paper presents a new high resolution aerial images dataset in which moving objects are labelled manually. It aims to contribute to the evaluation of the moving object detection methods for moving cameras. The problem of recognizing moving objects from aerial images is one of the important issues in computer vision. The biggest problem in the images taken by UAV is that the background is constantly variable due to camera movement. There are various datasets in the literature in which proposed methods for motion detection are evaluated. Prepared dataset consists of challenging images containing small targets compared to other datasets. Two methods in the literature have been tested for the prepared dataset. In addition, a simpler method compared to these methods has been proposed for moving object object in this paper.