Monitoring forest areas is of paramount importance to maintain environmental sustainability. The scalability of forest monitoring solutions is effectively offered by satellite imaging, where images of various modalities are acquired in orbit and cover large areas. However, building machine learning models for such downstream Earth observation tasks is challenging due to the limited amounts of ground-truth datasets. We tackle this issue and introduce an end-to-end deep learning pipeline to detect forest changes from Sentinel-2 time series of multispectral images. It benefits from a foundation model fine-tuned over a small yet spatially diverse dataset. The experiments showed that not only does it outperform other deep models, but also it requires minimal user intervention before the fine-tuning process.
Single-image super-resolution refers to the reconstruction of a high-resolution image from a single low-resolution observation. Although recent deep learning-based methods have demonstrated notable success on simulated datasets—with low-resolution images obtained by degrading and downsampling high-resolution ones—they frequently fail to generalize to real-world settings, such as document scans, which are affected by complex degradations and semantic variability. In this study, we introduce a task-driven, multi-task learning framework for training a super-resolution network specifically optimized for optical character recognition tasks. We propose to incorporate auxiliary loss functions derived from high-level vision tasks, including text detection using the connectionist text proposal network (CTPN), text recognition via a convolutional recurrent neural network (CRNN), keypoints localization using Key.Net, and hue consistency. To balance these diverse objectives, we employ a dynamic weight averaging (DWA) mechanism, which adaptively adjusts the relative importance of each loss term based on its convergence behavior. Experimental evaluation demonstrates that the proposed approach improves text detection, measured with intersection over union, by 1.09% for simulated and 1.94% for real-world datasets containing scanned documents, while preserving overall image fidelity. These improvements are statistically significant as confirmed by the Kruskal–Wallis H test and the post hoc Dunn test with Benjamini–Hochberg p-value correction. Our findings highlight the value of multi-objective optimization in super-resolution models for bridging the gap between simulated training regimes and practical deployment in real-world scenarios.
Selecting an appropriate neural network architecture and hyper-parameters to optimize performance for a given application is a difficult task. To overcome this challenge, Neural Architecture Search (NAS) and Hyperparameter Optimization (HPO) have been introduced. However, these techniques are often computationally-expensive and require significant amounts of execution time. To address this issue, we propose a semi-automated NAS approach that optimizes pre-existing architectural structures using genetic algorithms and eliminates unsuccessful combinations through shallow network training. The effectiveness of our technique was verified through an experiment that produces a family of AlexNet-like neural networks comprising 1296 models for image classification tasks. The computational study was conducted with five runs of a genetic algorithm, resulting in the deep models with mean loss of 0.7044 and accuracy of 0.7334, both with low standard deviations, outperforming the original AlexNet model with a significant margin.
15 October 2025 in Annals of Computer Science and Information Systems
Assessing the quality of super-resolved images is important for comparing reconstruction algorithms, but pixel fidelity, perceptual appearance, and structural preservation do not always agree. We investigate whether keypoint detector response maps and detected keypoints can act as trainable structural indicators for aligned full-reference super-resolution image quality assessment (SR-IQA). A contrastive Multi-Scale Index Proposal (MSIP) objective specializes Key.Net toward SR-like resolution loss, candidate checkpoints are screened without subjective labels, and six response- and keypoint-based measures are evaluated on four subjective benchmarks. The results show a redistribution rather than a uniform improvement: MSIP correlations increase on three of the four benchmarks, both repeatability variants decrease for every trained family on SISAR and RealSRQ, and general keypoint performance on HPatches decreases for trained checkpoints. At the benchmark level, established comparators such as TOPIQ-FR, RQI, and DISQ remain stronger, and on three of the four datasets the best keypoint-based result is still obtained with the pretrained detector. The measures nevertheless retain quality-related variation after conditioning on ten IQA controls in 88 of 120 tested hypotheses, with markedly weaker evidence on RealSRQ. Their practical value is therefore diagnostic: the response maps localize the structures behind an HR–SR discrepancy, complementing rather than replacing established SR-IQA metrics.