The field of computer vision is moving towards the development of efficient models that can perform real-time detection and segmentation tasks. Researchers are focusing on designing lightweight architectures that can achieve state-of-the-art performance while reducing computational complexity. This is particularly important for applications such as fire detection, wildlife monitoring, and disaster response, where timely and accurate detection is crucial.

Noteworthy papers include: Light-YOLOv8-Flame, which achieves a 0.78% gain in mean average precision and a 2.05% boost in recall for flame detection. LightFormer, a lightweight decoder for remote sensing image segmentation that achieves 99.9% of GLFFNet's mIoU while requiring only 14.7% of its FLOPs and 15.9% of its parameters. Change State Space Models for Remote Sensing Change Detection, which reduces the number of network parameters and enhances computational efficiency while maintaining high detection performance. CFIS-YOLO, a lightweight object detection model optimized for edge devices that achieves a mean Average Precision of 77.5% for wood defect detection. Post-Hurricane Debris Segmentation Using Fine-Tuned Foundational Vision Models, which fine-tunes pre-trained foundational vision models to achieve robust performance with a relatively small dataset. SAM-Based Building Change Detection with Distribution-Aware Fourier Adaptation and Edge-Constrained Warping, which proposes a new network with Distribution-Aware Fourier Aggregated Adapter to aggregate task-oriented changed information. Stronger, Steadier & Superior: Geometric Consistency in Depth VFM Forges Domain Generalized Semantic Segmentation, which integrates depth information with features from Vision Foundation Models to improve geometric consistency and boost generalization performance. RF-DETR Object Detection vs YOLOv12, which compares the performance of RF-DETR and YOLOv12 for greenfruit detection in complex orchard environments.