From: Object detection for automotive radar point clouds – a comparison
Method | IOU=0.3 | IOU=0.5 | ||||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
APped | APgrp | APcyc | APcar | APtrk | mAP | mLAMR | F1,pt | F1,obj | APped | APgrp | APcyc | APcar | APtrk | mAP | mLAMR | F1,pt | F1,obj | |
Metric | ||||||||||||||||||
DBSCAN/LSTM | 22.96 | 42.95 | 62.51 | 65.21 | 59.07 | 50.54 | 53.10 | 53.06 | 58.92 | 22.94 | 41.07 | 62.51 | 63.13 | 56.36 | 49.20 | 54.90 | 53.08 | 57.60 |
DBSCAN/RandomForest | 20.67 | 37.28 | 53.98 | 59.63 | 56.53 | 45.62 | 59.68 | 49.48 | 52.43 | 20.60 | 35.23 | 53.76 | 55.91 | 50.97 | 43.29 | 61.58 | 49.50 | 51.10 |
PointNet++/DBSCAN | 29.42 | 53.06 | 56.37 | 53.15 | 26.19 | 43.64 | 61.10 | 54.55 | 51.67 | 27.47 | 47.56 | 54.85 | 49.16 | 21.10 | 40.03 | 64.17 | 54.37 | 49.24 |
YOLOv3 | 28.28 | 57.51 | 64.87 | 75.54 | 62.18 | 57.67 | 48.92 | 53.04 | 61.87 | 26.96 | 54.88 | 63.68 | 67.99 | 56.31 | 53.96 | 52.61 | 52.86 | 59.46 |
PointPillars | 15.13 | 32.28 | 46.40 | 61.19 | 47.48 | 40.50 | 65.46 | 47.62 | 45.92 | 14.67 | 30.42 | 45.08 | 54.38 | 39.87 | 36.89 | 68.55 | 47.70 | 43.75 |
PointPillars++ | 24.69 | 47.32 | 55.39 | 68.72 | 53.08 | 49.84 | 57.01 | 53.91 | 55.61 | 23.93 | 44.66 | 53.58 | 61.33 | 45.62 | 45.82 | 60.36 | 53.93 | 53.36 |
PointNet++/LSTM | 27.28 | 56.28 | 64.94 | 68.44 | 60.21 | 55.43 | 50.10 | 54.13 | 60.99 | 27.17 | 51.70 | 64.88 | 63.74 | 57.00 | 52.90 | 52.06 | 54.12 | 59.64 |