Sill-Net: Feature Augmentation with Separated Illumination Representation

Benchmark Model Rank Results
traffic-sign-recognition-on-gtsrbSill-Net#2Accuracy: 99.68%
few-shot-image-classification-on-cifar-fs-5-way-1-shotIllumination Augmentation#5Accuracy: 87.73
few-shot-image-classification-on-cub-200-5-way-5-shotIllumination Augmentation#5Accuracy: 96.28
few-shot-image-classification-on-cifar-fs-5-way-5-shotIllumination Augmentation#6Accuracy: 91.09
few-shot-image-classification-on-cub-200-5-way-1-shotIllumination Augmentation#6Accuracy: 94.73
traffic-sign-recognition-on-tsinghua-tencent-100kSill-Net#6Accuracy: 99.53
few-shot-image-classification-on-mini-imagenet-5-way-1-shotIllumination Augmentation#10Accuracy: 82.99
few-shot-image-classification-on-mini-imagenet-5-way-5-shotIllumination Augmentation#13Accuracy: 89.14