跳到正文
RCreddit.com·

Automotive Radar Object Classification [P]

AI 摘要

一项针对汽车雷达目标分类的模型研究显示,随着每个实例的雷达检测数量增加,模型性能有所提升。当检测数量从1个增加到5个时,宏观F1分数从0.381上升至0.764。该模型使用所有可用检测进行训练,然后根据每个实例的检测计数对现有验证预测进行分组,并计算每个组的宏观F1分数。未来的工作包括实施其他空间编码方案(例如点云网络)并累积多次扫描,以解决稀疏性挑战并探索微多普勒概念。

时间与来源
发布
2026年9月7日 08:10
来源类型
开发者社区
档位
社区
信源状态
正常
档位是按信源手工设定的编辑判断,不是逐条打分。

时间以 UTC 显示

更多信息
首次发现2026年9月7日 10:00时区UTC · UTC+0
正文

I'm a radar signal processing engineer and i trained a 5-class classifier (car, large_vehicle, two_wheeler, pedestrian, pedestrian_group) on RadarScenes radar point clouds.

The input vector is a per-scan histogram (16 bins) and the network is a 3-layer MLP. The loss function is a class-weighted cross-entropy loss. This work is based on "Histogram-based Deep Learning for Automotive Radar" paper.

I scoped the project to be one scan only. Accumulation of multiple scans is the next step.

Aggregated Classes: two_wheeler mixes bicycles and motorized variants; large_vehicle merges trucks, buses, and trains together due to data scarcity.

Sequence Bias: Long tracks of slow-moving objects can skew a particular data split velocity distribution, causing high F1 score variance across folds.

I tried with bigger MLPs, alternative feature encodings, and different histogram binning, all moved performance less than the variation caused by changing the train/validation/test split. I measured that split sensitivity across 6 folds, keeping the same proportions.

Macro F1 rises from 0.381 to 0.764 as the naturally occurring number of radar detections per instance increases from 1 to 5. I trained the model normally using all available detections, then bucketed its existing validation predictions by each instance's detection count and computed macro F1 per bucket.

The classes car and pedestrian has the best performance and two_wheeler has the worst.

A car is often confused as large vehicle when the car was wider than usual or had a unusually high rcs (which can happen due to multipath for example).

The two_wheeler is often confused as pedestrian because their vr_compensated distributions overlap, which is the the model's single most important feature for these two classes. A stationary or idling two_wheeler is indistinguishable from a pedestrian.

I uploaded an image with ground truth vs predictions: A nearly stationary two-wheeler which contains a single point was predicted as pedestrian, because its velocity is near zero, indistinguishable from a pedestrian. A car in the same scene, also with just one point, is classified correctly, since RCS and Doppler are enough for that class.

Full writeup here: https://github.com/brunopinto900/radar-ml-autonomous-driving/blob/main/MLP_Report.md

Implement other spatial encoding schemas (point net for example) and accumulate multiple scans to tackle the challenge of sparsity and explore the concept of micro-doppler.

来源·reddit.com·RSS 全文