Table 1.
Performance of different algorithm on Cornell Grasp Dataset.
| Approach | Algorithm | Accuracy (%) | |
|---|---|---|---|
| Image-Wise | Object-Wise | ||
| Jiang et al. [1] | Fast Search | 60.5 | 58.3 |
| Lenz et al. [2] | SAE, struct. reg. Two stage | 73.9 | 75.6 |
| Redmon et al. [3] | AlexNet, MultiGrasp | 88.0 | 87.1 |
| Kumra et al. [4] | ResNet-50 × 2, Multi-model Grasp Predictor | 89.21 | 88.96 |
| Guo et al. [5] | ZF-net, Hybrid network, 3 scales and 3 aspect ratios | 93.2 | 89.1 |
| Chu et al. [11] | VGG-16 model | 95.5 | 91.7 |
| ResNet-50 model | 96.0 | 96.1 | |
| Zhou et al. [6] | ResNet-50 FCGN | 97.7 | 94.9 |
| ResNet-101 FCGN | 97.7 | 96.6 | |
| the proposed scheme | keypoint-based scheme | 96.05 | 96.5 |