Euijoon Ahn

Foundation models, pretrained on diverse data, objectives, or modalities for reuse across downstream tasks, are increasingly applied to medical image analysis. Their performance depends not only on pretraining but also on how they are adapted to specific tasks, data regimes, and deployment settings. Whereas previous reviews have considered adaptation within broader syntheses, this review focuses specifically on post-pretraining interventions applied to existing FMs. Using a structured narrative review with a systematic, auditable workflow, we identified 719 eligible reports representing 705 independent studies. Of these, 473 provided controlled comparative evidence: 322 formally published studies formed the primary synthesis and 151 preprint-only studies a sensitivity layer. Adaptation was organized into five non-mutually exclusive families: parameter-level, representation-level, objective-level, data-centric, and architectural/sequence-level adaptation, with model use and update regime analyzed separately. Controlled comparisons showed no consistent ranking among frozen, parameter-efficient, partial, or full updating, or across broader adaptation families; outcomes varied by task, data regime, comparator, and evaluation setting. Clinical-readiness evidence remained limited: 173/705 studies (24.5%) had confirmed independent external validation, 10 (1.4%) reported calibration or uncertainty evidence, and 97/322 formally published comparative studies (30.1%) included a qualifying externally validated arm. Overall, FM adaptation is a context-dependent design space rather than a hierarchy of preferred strategies.

Passive Acoustic Monitoring will transform biodiversity assessment and large-scale ecological surveys through continuous ecoacoustic data collection. Yet, dataset growth has outpaced existing segmentation methods, many of which are tuned to narrow taxonomic subsets lacking abiotic sounds, limiting ecological realism. Natural soundscapes contain overlapping biophony, geophony, and anthrophony, making scalable and ecologically valid segmentation an ongoing challenge. As such, we introduce an unsupervised framework designed to reveal meaningful acoustic structures without predefined labels. The pipeline integrates systematic sampling, sound event detection, Mel-Frequency Cepstral Coefficient extraction, dimensionality reduction via Uniform Manifold Approximation and Projection (UMAP), and density-based clustering with HDBSCAN. We evaluate the framework across six biodiverse Australian soundscapes comprising of (n=19,230) 4.5-second non-overlapping segments. Results indicate that clusters are internally coherent, with Voronoi tessellations over the UMAP space showing distinct spatial boundaries. External validation with (n=2,000) manually annotated samples demonstrates strong alignment with ecologically meaningful sound types. F1-scores ranged from 83.7% to 98.3%, with precision and recall exceeding 91% across most sites. By combining unsupervised clustering with ecological validation, our framework offers a scalable, generalisable solution for organising unlabelled ecoacoustic data while reducing manual effort and preserving ecological integrity.

COVID-19 pandemic has greatly impacted travel and hospitality industry. As hotels began to reopen after the initial lockdowns, many implemented new standards to reduce health risks. At the same time, travel recommender sites and accommodation booking websites introduced sustainability labelling and health and safety badges to help travellers book safe and sustainable accommodation options. This study investigates a 2,655 online hotel reviews from Costa del Sol region of Spain, collected between February 2020 and March 2021 to capture the full impact of the pandemic’s first year on tourism sector. Using a deep learning-based sentiment analysis combined with ANOVA, the study examines gest perceptions of hotels’ health and sustainability standards and explores their relationship with overall review scores. Additionally, the content of the reviews is analysed to identify emerging themes and associated sentiments. The findings indicate that although guests rarely mentioned health measures explicitly, these practices had a measurable impact on review scores. While the analysis of sustainability-related sentiments remains exploratory, preliminary results suggest that higher-category hotels (i.e., 4- or 5-star) are more likely to communicate a broader range of implemented sustainability, health, and safety practices. More research is needed to understand the evolving sentiment towards health and sustainability issues amongst hotel guests across various segments of travel and hospitality industry as wells as across cultures.

In industrial automation, radar is a critical sensor in machine perception. However, the angular resolution of radar is inherently limited by the Rayleigh criterion, which depends on both the radar's operating wavelength and the effective aperture of its antenna array.To overcome these hardware-imposed limitations, recent neural network-based methods have leveraged high-resolution LiDAR data, paired with radar measurements, during training to enhance radar point cloud resolution. While effective, these approaches require extensive paired datasets, which are costly to acquire and prone to calibration error. These challenges motivate the need for methods that can improve radar resolution without relying on paired high-resolution ground-truth data. Here, we introduce an unsupervised radar points enhancement algorithm that employs an arbitrary LiDAR-guided diffusion model as a prior without the need for paired training data. Specifically, our approach formulates radar angle estimation recovery as an inverse problem and incorporates prior knowledge through a diffusion model with arbitrary LiDAR domain knowledge. Experimental results demonstrate that our method attains high fidelity and low noise performance compared to traditional regularization techniques. Additionally, compared to paired training methods, it not only achieves comparable performance but also offers improved generalization capability. To our knowledge, this is the first approach that enhances radar points output by integrating prior knowledge via a diffusion model rather than relying on paired training data. Our code is available at https://github.com/yyxr75/RadarINV.