The field of computer vision is witnessing significant advancements in cross-domain few-shot object detection and image restoration. Researchers are developing innovative models that can efficiently detect objects across different domains with limited labeled data. Notably, the focus is on designing models that can optimize key computational metrics such as runtime, parameters, and FLOPs while achieving state-of-the-art results. The use of event-based vision and dual-focused images is also being explored for tasks like image deblurring and raindrop removal. These advancements have the potential to drive further research in the field and establish new benchmarks for future studies. Noteworthy papers include the NTIRE 2025 Challenge on Cross-Domain Few-Shot Object Detection, which proposed novel models achieving new state-of-the-art results. The NTIRE 2025 Challenge on Event-Based Image Deblurring and the NTIRE 2025 Challenge on Day and Night Raindrop Removal for Dual-Focused Images also presented innovative approaches and established benchmarks for future research.