Войти
аналитика - 3D object detection
Unsupervised 3D Object Detection for Roadside Units: A Label-Efficient Approach
We present a label-efficient 3D object detection method for roadside units based on unsupervised object discovery and refinement, which can achieve comparable performance to fully supervised models with only a small amount of manually labeled data.
Unified Open-Vocabulary 3D Object Detection via Cycle-Modality Propagation
OV-Uni3DETR, a unified open-vocabulary 3D detector, leverages multi-modal data and cycle-modality propagation to enable open-vocabulary, modality-switchable, and scene-unified 3D object detection.