Author Affiliations
1School of Electrical and Electronic Engineering, Shanghai University of Engineering Science, Shanghai 201620, China2School of Information Science and Engineering, Wuhan University of Science and Technology, Wuhan 430081, Hubei , Chinashow less
Fig. 1. Structure of IE-Conv
Fig. 2. Schematic diagrams of inner and outer point set module. (a) Internal point set module; (b) external point set module
Fig. 3. Structure of internal point set module
Fig. 4. Structure of external point set module
Fig. 5. Architecture of interior-exterior point set shape feature convolutional network
Method | Input | Points /103 | Acc /% |
---|
Pointwise-CNN[31] | pnt | 1 | 86.1 | ECC[19] | pnt | 1 | 87.4 | PointNet[8] | pnt | 1 | 89.2 | Point-CNN[33] | pnt | 1 | 91.7 | DGCNN[10] | pnt | 1 | 92.2 | SO-CNN[15] | pnt | 1 | 93.1 | Dense-Point[26] | pnt | 1 | 93.2 | RS-CNN[14] | nor | 1 | 92.8 | PAT[27] | pnt,nor | 1 | 91.7 | Spec-GCN[30] | pnt,nor | 1 | 91.8 | PointConv[1] | pnt,nor | 1 | 92.5 | A-CNN[11] | pnt,nor | 1 | 92.6 | PointASNL[13] | pnt,nor | 1 | 93.2 | ELM[28] | pnt,nor | 1 | 93.2 | RS-CNN[14] | pnt,nor | 1 | 93.6 | SO-Net[29] | pnt,nor | 2 | 90.9 | PointNet++[9] | pnt,nor | 5 | 91.9 | Spider-CNN[32] | pnt,nor | 5 | 92.4 | SO-Net[29] | pnt,nor | 5 | 93.4 | Proposed method | pnt,nor | 1 | 93.9 |
|
Table 1. Comparison of classification accuracy for ModelNet40 dataset
Method | Number of parameters /MB | Acc /% | FLOPs /sample |
---|
PointNet | 3.50 | 89.2 | 440 | Spec-GCN | 2.05 | 91.8 | 1112 | PointNet++ | 1.48 | 91.9 | 1684 | DGCNN | 1.84 | 92.2 | 2767 | Proposed method | 1.37 | 93.9 | 266 |
|
Table 2. Comparison of classification complexity and time of the ModelNet40 dataset
Method | Air | Bag | Cap | Car | Chai | Ear. | Gui. | Knife | Lamp | Lap | Moto | Mug | Pistol | Rock | Skate | Table | Mean |
---|
PointNet[8] | 83.4 | 78.7 | 82.5 | 74.9 | 89.6 | 73.0 | 91.5 | 85.9 | 80.8 | 95.3 | 65.2 | 93.0 | 81.2 | 57.9 | 72.8 | 80.6 | 83.7 | SONet[29] | 82.8 | 77.8 | 88.0 | 77.3 | 90.6 | 73.5 | 90.7 | 83.9 | 82.8 | 94.8 | 69.1 | 94.2 | 80.9 | 53.1 | 72.9 | 83.0 | 84.9 | PointNet++[9] | 82.4 | 79.0 | 87.7 | 77.3 | 90.8 | 71.8 | 91.0 | 85.9 | 83.7 | 95.3 | 71.6 | 94.1 | 81.3 | 58.7 | 76.4 | 82.6 | 85.1 | DGCNN[10] | 84.2 | 83.7 | 84.4 | 77.1 | 90.9 | 78.5 | 91.5 | 87.3 | 82.9 | 96.0 | 67.8 | 93.3 | 82.6 | 59.7 | 75.5 | 80.6 | 85.1 | PCNN[1] | 82.4 | 80.1 | 85.5 | 79.5 | 90.8 | 73.2 | 91.3 | 86.0 | 85.0 | 95.7 | 73.2 | 94.8 | 83.3 | 51.0 | 75.0 | 81.8 | 85.1 | ELM[28] | 84.0 | 80.4 | 88.0 | 80.2 | 90.7 | 77.5 | 91.2 | 86.4 | 82.6 | 95.5 | 70.0 | 93.9 | 84.1 | 55.6 | 75.6 | 82.1 | 85.3 | SpiderCNN[32] | 83.5 | 81.0 | 87.2 | 77.5 | 90.7 | 76.8 | 91.1 | 87.3 | 83.3 | 95.8 | 70.2 | 93.5 | 82.7 | 59.7 | 75.8 | 82.8 | 85.3 | SO-CNN | 83.9 | 84.1 | 85.0 | 77.4 | 91.3 | 78.3 | 91.7 | 87.4 | 83.8 | 96.4 | 69.7 | 93.5 | 83.1 | 58.9 | 76.2 | 82.9 | 85.7 | A-CNN[11] | 84.2 | 84.0 | 88.0 | 79.6 | 91.3 | 75.2 | 91.6 | 87.1 | 85.5 | 95.4 | 75.3 | 94.9 | 82.5 | 67.8 | 77.5 | 83.3 | 86.1 | RS-CNN[14] | 83.5 | 84.8 | 88.8 | 79.6 | 91.2 | 81.1 | 91.6 | 88.4 | 86.0 | 96.0 | 73.7 | 94.1 | 83.4 | 60.5 | 77.7 | 83.6 | 86.2 | Proposed method | 84.0 | 86.2 | 88.1 | 79.5 | 91.6 | 77.5 | 91.3 | 88.0 | 86.3 | 96.1 | 72.8 | 95.0 | 83.6 | 62.2 | 75.9 | 83.9 | 86.4 |
|
Table 3. Comparison of segmentation accuracy of ShapeNet dataset
Object | Visualization | Object | Visualization |
---|
Air | | | Lamp | | | Bag | | | Laptop | | | Cap | | | Motor. | | | Car | | | Mug | | | Chair | | | Pistol | | | Ear. | | | Rocket | | | Gui. | | | Skate | | | Knife | | | Table | | |
|
Table 4. Visualization results of SR-Net) on ShapeNet dataset
Model | RS-CNN(*) | Internal | External | Internal-external | Acc /% |
---|
A | | | | | 90.1 | B | √ | | | | 92.0 | C | | √ | | | 93.2 | D | √ | | √ | | 92.9 | E | | | | √ | 93.9 |
|
Table 5. Ablation experiments with modules on ModelNet40 dataset
Model | Priori expressions | Channel | Acc /% |
---|
A | | 1 | 92.4 | | 1 | B | 、 | 6 | 93.4 | 、 | 6 | C | 、、 | 7 | 93.9 | 、、 | 7 | D | 、、 | 7 | 93.0 | E | 、、 | 7 | 92.5 |
|
Table 6. Ablation experiments on ModelNet40 dataset for different prior expressions as gates
Method | Self-calibrate | Translation | Rotate |
---|
-0.2 | +0.2 | 90˚ | 180˚ |
---|
PointNet | | 70.8 | 70.6 | 42.5 | 38.6 | PointNet++ | | 88.2 | 88.2 | 47.9 | 39.7 | Proposed method | | 90.9 | 90.9 | 90.9 | 90.9 | Proposed method | √ | 92.1 | 92.1 | 92.1 | 92.1 |
|
Table 7. Robustness experiments on ModelNet40 dataset after adding translations or rotations to point clouds
Model | Aggregation function | Self-calibrate | Acc /% |
---|
A | Avg | | 92.7 | B | Max | | 93.2 | C | Max | √ | 93.5 | D | Sum | | 93.5 | E | Sum | √ | 93.9 |
|
Table 8. Ablation experiments of different aggregation functions and self calibration functions on ModelNet40 dataset