Grasp Detection on Cornell Grasping Dataset (Image-wise split)
89.21Detection AccuracyMulti-Modal Grasp Predictor
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Multi-Modal Grasp PredictorBackbone=ResNet-50x2, Activation=ReLU, ReLU, ReLU, Input=RGB-D2016.11 | 89.21 | — | — | |
| Uni-modal Grasp PredictorBackbone=ResNet-50, Activation=ReLU, ReLU, Input=RGB2016.11 | 88.84 | — | — | |
| Uni-modal Grasp PredictorBackbone=ResNet-50, Activation=tanh, ReLU, Input=RGD2016.11 | 88.53 | — | — | |
| STEM-CaRFs (Selective Grasp)2016.11 | 88.2 | — | — | |
| MultiGrasp Detection2014.12 | 88 | — | — | |
| AlexNet, MultiGrasp2016.11 | 88 | — | — | |
| Multi-Modal Grasp PredictorBackbone=ResNet-50x2, Classifier=linear SVM, Input=RGB-D2016.11 | 86.44 | — | — | |
| Regression + ClassificationTime / image=76 ms2014.12 | 85.5 | — | — | |
| Two-stage closed-loop, with penalty2016.11 | 85.3 | — | — | |
| Uni-modal Grasp PredictorBackbone=ResNet-50, Classifier=SVM, Input=RGB, Note=Baseline2016.11 | 84.76 | — | — | |
| Direct Regression2014.12 | 84.4 | — | — | |
| Lenz et al.Time / image=13.5 sec2014.12 | 73.9 | — | — | |
| SAE, struct. reg. two-stage2016.11 | 73.9 | — | — | |
| Jiang et al.2014.12 | 60.5 | — | — | |
| Fast Search2016.11 | 60.5 | — | — | |
| Chance2014.12 | 6.7 | — | — | |
| Chance2016.11 | 6.7 | — | — | |
| Chance2013.01 | — | 35.9 | 6.7 | |
| Jiang et al.2013.01 | — | 75.3 | 60.5 | |
| SAE, L1 reg.Regularization=L12013.01 | — | 87.2 | 72.9 | |
| SAE, no mask-based scalingScaling=No mask-based2013.01 | — | 62.1 | 39.9 | |
| SAE, separate layer-1 feat.Features=Separate layer-12013.01 | — | 70.3 | 43.3 | |
| SAE, struct. reg. two-stageRegularization=Structured (Multimodal Group), Architecture=Two-stage cascaded2013.01 | — | 88.4 | 73.9 | |
| SAE, struct. reg., 1st pass onlyRegularization=Structured (Multimodal Group), Pass=1st only2013.01 | — | 86.4 | 70.6 | |
| SAE, struct. reg., 2nd pass onlyRegularization=Structured (Multimodal Group), Pass=2nd only2013.01 | — | 87.5 | 73.8 |