In this report, we introduce our winning solution to the Real-time 3D Detection and also the "Most Efficient Model" in the Waymo Open Dataset Challenges at CVPR 2021. Extended from our last year's award-winning model AFDet, we have made a handful of modifications to the base model, to improve the accuracy and at the same time to greatly reduce the latency. The modified model, named as AFDetV2, is featured with a lite 3D Feature Extractor, an improved RPN with extended receptive field and an added sub-head that produces an IoU-aware confidence score. These model enhancements, together with enriched data augmentation, stochastic weights averaging, and a GPU-based implementation of voxelization, lead to a winning accuracy of 73.12 mAPH/L2 for our AFDetV2 with a latency of 60.06 ms, and an accuracy of 72.57 mAPH/L2 for our AFDetV2-base, entitled as the "Most Efficient Model" by the challenge sponsor, with a winning latency of 55.86 ms.

中文翻译：

---

具有 IoU 意识的实时无锚单阶段 3D 检测

在这份报告中，我们介绍了我们在 2021 年 CVPR 上的 Waymo 开放数据集挑战赛中实时 3D 检测的获奖解决方案以及“最高效模型”。从我们去年的获奖模型 AFDet 扩展而来，我们制作了一些对基础模型的修改，以提高准确性，同时大大减少延迟。修改后的模型命名为 AFDetV2，具有 lite 3D 特征提取器、具有扩展感受野的改进 RPN 和增加的子头，可产生 I​​oU 感知置信度分数。这些模型增强，连同丰富的数据增强、随机权重平均和基于 GPU 的体素化实现，使我们的 AFDetV2 的获胜精度为 73.12 mAPH/L2，延迟为 60.06 毫秒，精度为 72.57 mAPH/ L2 用于我们的 AFDetV2-base，