karthik789338/OccDet-Calib

OccDet-Calib是一个用于研究部分可见性下检测器校准的基准和分析框架，包含多个子集和条件，如重叠遮挡、干扰物控制、截断控制和修复鲁棒性子集，以及基于BDD100K风格注释的自然验证。主要类别包括人、自行车、摩托车、汽车、公共汽车、卡车和停车标志。数据集提供基准元数据CSV、处理后的评估表、论文图表和样本资源，但不包含原始图像或检查点。

OccDet-Calib is a benchmark and analysis framework for studying detector calibration under partial visibility, including multiple subsets and conditions such as overlap occlusion, distractor control, truncation control, inpainting robustness subset, and natural validation on BDD100K-style annotations. Main classes include person, bicycle, motorcycle, car, bus, truck, and stop sign. The dataset provides benchmark metadata CSVs, processed evaluation tables, paper figures, and sample assets, but does not include raw images or checkpoints.