Xiongwen Luo

Title: Xiongwen Luo and Scene Recognition Model Training

Introduction

Xiongwen Luo is an inventor associated with Guangzhou, China, and Bigo Technology Pte. Ltd. His patent work includes a method for training a scene recognition model, combining feature extraction, attention-based processing, and classification.

Latest Patents

Luo is named on 1 patent: “Model Training and Scene Recognition Method and Apparatus, Device, and Medium.” The invention describes training a scene recognition model with a core feature extraction layer, a global information feature extraction layer, one or more attention-based LCS modules, and a fully connected decision layer.

The method trains the core and global feature extraction layers using a sample image’s scene label and standard cross-entropy loss. It trains the weight parameters of each LCS module using pixel-by-pixel loss calculations between the module’s output feature map and the scene label. The fully connected decision layer is then trained using the scene label and standard cross-entropy loss.

Career Highlights

The available information connects Luo’s patent activity with Bigo Technology Pte. Ltd. His listed invention focuses on a structured approach to training models for scene recognition.

Collaborations

Luo’s listed coworkers include Jianghu Lu and Wei Xiang. The provided information does not specify their roles in the patent or the extent of their collaboration.

Conclusion

Xiongwen Luo’s recorded patent contribution concerns a scene recognition model training method that combines layered feature extraction, attention-based modules, and a decision layer.