The field of computer vision is rapidly advancing, with a focus on improving image processing and 3D modeling techniques. Recent research has led to the development of innovative methods for defect detection, feature matching, and semantic correspondence. These advancements have the potential to significantly impact various applications, including quality control, robotics, and urban planning. Notably, deep learning-based approaches have shown exceptional performance in tasks such as image registration, structure-from-motion, and visual localization. Furthermore, the use of wavelet-based frameworks and graph neural networks has demonstrated promising results in stereo matching and loop closure detection. Overall, the field is moving towards more efficient and accurate methods for image processing and 3D modeling, with a focus on real-world applications. Noteworthy papers include:

• A research paper that proposes a wavelet-based stereo matching framework, which achieves state-of-the-art performance on benchmark datasets.
• Another paper that introduces a deep learning-based approach for building floor number estimation from crowdsourced street-level images, achieving high accuracy and scalability.
• A study that presents a comprehensive survey of semantic correspondence methods, providing a unified benchmark and a strong baseline for future research.