This is a page that curates AI-related papers published worldwide. All content here is summarized using Google Gemini and operated on a non-profit basis. Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.
UniOcc is a comprehensive, integrated benchmark and toolkit for occupancy prediction (predicting future occupancy based on historical information) and occupancy forecasting (predicting current frame occupancy based on camera images). It integrates real-world datasets like nuScenes and Waymo with data from high-fidelity driving simulators like CARLA and OpenCOOD to provide 2D/3D occupancy labels and innovative pixel-by-pixel flow annotation. By avoiding the suboptimal pseudo-labels used in existing studies and introducing novel evaluation metrics that do not rely on true labels, it enables robust assessment of additional aspects of occupancy quality. Extensive experiments with state-of-the-art models demonstrate that large, diverse training data and explicit flow information significantly improve occupancy prediction and forecast performance. Data and code are available at https://uniocc.github.io/ .
Contributes to occupancy prediction and improved prediction performance by integrating large and diverse datasets.
◦
A more robust evaluation is possible by presenting a new evaluation metric that does not rely on the correct answer label.
◦
Improve performance by leveraging per-pixel flow information.
◦
Research reproducibility and advancement are possible through open datasets and code.
•
Limitations:
◦
The currently proposed Limitations is not explicitly mentioned in the paper. Further research is needed to evaluate the algorithm's generalization performance and vulnerability to specific environments.
