Table 7 Comparison with several popular methods on KITTI dataset (%)
From: Multiclass objects detection algorithm using DarkNet-53 and DenseNet for intelligent vehicles
Method | Data | Car | Pedestrian | Cyclist | ||||||
|---|---|---|---|---|---|---|---|---|---|---|
Easy | Moderate | Hard | Easy | Moderate | Hard | Easy | Moderate | Hard | ||
Faster R-CNN [14] | Image | 88.97 | 83.16 | 72.62 | 79.97 | 66.24 | 61.09 | 72.40 | 62.86 | 54.97 |
YOLOv3 [9] | Image | 88.71 | 74.40 | 65.58 | 67.23 | 49.47 | 44.99 | 50.88 | 36.89 | 32.64 |
YOLOv5X6 | Image | 96.64 | 93.82 | 81.54 | 81.88 | 64.53 | 57.44 | 75.21 | 52.99 | 45.67 |
Stereo R-CNN [44] | Stereo Image | 93.98 | 85.98 | 71.25 | – | – | – | – | – | – |
3DOP [45] | Stereo Image | 92.96 | 89.55 | 79.38 | 83.17 | 69.57 | 63.48 | 80.52 | 68.71 | 61.07 |
MV3D [46] | Image Lidar | 96.47 | 90.83 | 78.63 | – | – | – | – | – | – |
F-PointNet [47] | Image Lidar | 95.85 | 95.17 | 85.42 | 89.83 | 80.13 | 75.05 | 86.86 | 73.16 | 65.21 |
Ours | Image | 92.05 | 86.89 | 80.52 | 82.91 | 67.37 | 58.45 | 75.36 | 58.97 | 46.31 |