Multi-view 3D Classification on 3D Warehouse
93.8Top-1 AccuracySplitNets
Evaluation results
| Method | Links | |
|---|---|---|
| SplitNetsBackbone=C, #Params=2.32M, #OPs=1.83G, Position=8-th (tot. 13) block, Feature size=8 * 14^2, Arch.=Concat.2022.04 | 93.8 | |
| SplitNetsBackbone=B, #Params=2.44M, #OPs=759M, Position=8-th (tot. 13) block, Feature size=8 * 12^2, Arch.=Concat.2022.04 | 93 | |
| SplitNetsBackbone=A, #Params=1.39M, #OPs=513M, Position=6-th (tot. 8) block, Feature size=16 * 6^2, Arch.=Concat.2022.04 | 92.3 | |
| MVCNN-suBackbone=EfficientNet-B0, #Params=5.29M, #OPs=4.62G, Position=Last Conv, Feature size=1280 * 7^2, Arch.=View-Pool2022.04 | 92.1 | |
| MVCNN-suBackbone=MNASNet-0.5, #Params=2.22M, #OPs=1.24G, Position=Last Conv, Feature size=1280 * 7^2, Arch.=View-Pool2022.04 | 91.4 | |
| Auto-MVCNNBackbone=AM-c36, #Params=4.70M, #OPs=6.90G, Position=Last Conv, Arch.=Searched2022.04 | 91 | |
| Auto-MVCNNBackbone=AM-c24, #Params=2.10M, #OPs=3.20G, Position=Last Conv, Arch.=Searched2022.04 | 90.5 | |
| SeqViewsBackbone=VGG-19, #Params=144M, #OPs=235 G, Position=Penultimate FC, Feature size=4096 * 1, Arch.=Bi-RNN2022.04 | 89.3 | |
| MV-LSTMBackbone=ResNet-18, #Params=11.2M, #OPs=21.9G, Position=Last Conv, Feature size=512 * 7^2, Arch.=Bi-LSTM2022.04 | 89.1 | |
| MVCNN-newBackbone=VGG-11, #Params=132 M, #OPs=90.0G, Position=Last Conv, Feature size=512 * 7^2, Arch.=View-Pool2022.04 | 88.7 | |
| MVCNN-suBackbone=MobileNet-v3, #Params=2.54M, #OPs=661M, Position=Last Conv, Feature size=576 * 7^2, Arch.=View-Pool2022.04 | 86 | |
| MVCNN-suBackbone=VGG-M, #Params=90.8M, #OPs=34.6G, Position=Last Conv, Feature size=512 * 13^2, Arch.=View-Pool2022.04 | 81.1 |