Inspur Information contributed two research papers to CVPR 2022. “CoDo: Contrastive Learning with Downstream Background Invariance for Detection” proposes a contrastive learning framework that improves consistency in image classification and object detection. The other thesis, “Scene Representation in Bird’s-Eye View from Surrounding Cameras with Transformers,” describes a machine learning model and framework used to transform camera image input to enhance environment perception in autonomous driving. 

2、Scene Representation in Bird’s-Eye View from Surrounding Cameras with Transformers.pdf