List your company
ResearchFormerly WizWand

Teachable Object Recognition on ORBIT CLU-VE 1.0 (test)

66.3Frame AccuracyProtoNets + LITE
49.6653.9858.362.62Jul 2, 2021
Updated 8mo ago

Evaluation results

MethodLinks
ProtoNets + LITEBackbone=EfficientNet-B0, Image Size=224, LITE Training=true, MACs to adapt=1.21T, Steps to adapt=1F, Wall clock time=1.72s, Number of parameters=4.01M2021.0766.312.772.9
Simple CNAPsBackbone=EfficientNet-B0, Image Size=224, LITE Training=false, MACs to adapt=3.39T, Steps to adapt=1F, Wall clock time=2.91s, Number of parameters=5.67M2021.0765.611.269.9
Simple CNAPs + LITEBackbone=EfficientNet-B0, Image Size=224, LITE Training=true, MACs to adapt=3.39T, Steps to adapt=1F, Wall clock time=2.92s, Number of parameters=5.67M2021.0765.613.571.9
MAMLBackbone=EfficientNet-B0, Image Size=224, LITE Training=false, MACs to adapt=36.31T, Steps to adapt=15FB, Wall clock time=117.89s, Number of parameters=4.01M2021.0764.612.869.4
CNAPsBackbone=EfficientNet-B0, Image Size=224, LITE Training=true, MACs to adapt=3.38T, Steps to adapt=1F, Wall clock time=2.85s, Number of parameters=10.59M2021.0763.312.869.2
FineTunerBackbone=EfficientNet-B0, Image Size=224, LITE Training=false, MACs to adapt=121.02T, Steps to adapt=50FB, Wall clock time=139.99s, Number of parameters=4.01M2021.0763.111.566.9
CNAPsBackbone=EfficientNet-B0, Image Size=224, LITE Training=false, MACs to adapt=3.38T, Steps to adapt=1F, Wall clock time=2.83s, Number of parameters=10.59M2021.0762.613.267.4
ProtoNets + LITEBackbone=ResNet-18, Image Size=224, LITE Training=true, MACs to adapt=5.47T, Steps to adapt=1F, Wall clock time=1.07s, Number of parameters=11.18M2021.0761.413.268.5
MAMLBackbone=ResNet-18, Image Size=224, LITE Training=false, MACs to adapt=163.97T, Steps to adapt=15FB, Wall clock time=65.22s, Number of parameters=11.18M2021.0759.316.364.3
CNAPsBackbone=ResNet-18, Image Size=224, LITE Training=true, MACs to adapt=7.64T, Steps to adapt=1F, Wall clock time=2.11s, Number of parameters=12.76M2021.0758.215.162.5
Simple CNAPsBackbone=ResNet-18, Image Size=224, LITE Training=false, MACs to adapt=7.64T, Steps to adapt=1F, Wall clock time=2.13s, Number of parameters=11.97M2021.0758.114.760.9
CNAPsBackbone=ResNet-18, Image Size=224, LITE Training=false, MACs to adapt=7.64T, Steps to adapt=1F, Wall clock time=2.11s, Number of parameters=12.76M2021.0757.614.966.5
Simple CNAPs + LITEBackbone=ResNet-18, Image Size=224, LITE Training=true, MACs to adapt=7.64T, Steps to adapt=1F, Wall clock time=2.14s, Number of parameters=11.97M2021.0757.517.364.6
ProtoNetsBackbone=EfficientNet-B0, Image Size=224, LITE Training=false, MACs to adapt=1.21T, Steps to adapt=1F, Wall clock time=1.72s, Number of parameters=4.01M2021.0757.312.763.9
ProtoNetsBackbone=ResNet-18, Image Size=224, LITE Training=false, MACs to adapt=5.47T, Steps to adapt=1F, Wall clock time=1.07s, Number of parameters=11.18M2021.0756.814.462.5
FineTunerBackbone=ResNet-18, Image Size=224, LITE Training=false, MACs to adapt=546.57T, Steps to adapt=50FB, Wall clock time=96.23s, Number of parameters=11.18M2021.0756.718.861.3
Simple CNAPsBackbone=ResNet-18, Image Size=84, LITE Training=false, MACs to adapt=3.48T, Steps to adapt=1F, Wall clock time=1.01s, Number of parameters=11.97M2021.0753.91662
FineTunerBackbone=ResNet-18, Image Size=84, LITE Training=false, MACs to adapt=317.70T, Steps to adapt=50FB, Wall clock time=53.94s, Number of parameters=11.17M2021.0753.714.463.1
MAMLBackbone=ResNet-18, Image Size=84, LITE Training=false, MACs to adapt=95.31T, Steps to adapt=15FB, Wall clock time=36.98s, Number of parameters=11.17M2021.0751.72157.9
CNAPsBackbone=ResNet-18, Image Size=84, LITE Training=false, MACs to adapt=3.48T, Steps to adapt=1F, Wall clock time=0.98s, Number of parameters=12.75M2021.0751.517.959.5
ProtoNetsBackbone=ResNet-18, Image Size=84, LITE Training=false, MACs to adapt=3.18T, Steps to adapt=1F, Wall clock time=0.73s, Number of parameters=11.17M2021.0750.314.959.9