Skip to main content

Table 1 Typical methods and elapsed times for each of the top models in the competition

From: Classification of imbalanced cloud image data using deep neural networks: performance improvement through a data science competition

  1st place model 2nd place model 3rd place model 4th place model Matsuoka et al. (2018)
CNN architecture PyramidNet WideResNet Shake-shake ResNet26 MobileNetV2 LeNet
Number of parameters 7.6 M 4.3 M 3.0 M 11.6 M 0.60 M
Preprocessing Binarization Downsampling (32 × 32) Upsampling (96 × 96)
Oversampling (Data augmentation) Vertical flip
Horizontal flip
Cutout
Random shift
Cropping
Random rotation
Random erasing
Random crop + Padding
Horizontal flip
Random rotation
Mixup
Undersampling Hard negative mining Random sampling Random sampling Random sampling Random sampling
Ensemble learning 5 models (Different hard negative ratio) 5 models (Different learning rate/preproccesing) 10 models (Different negative samples)
TTA Same as training phase 5 Crop × 4 Rotation 10 Crop + Padding
Others Focal loss RReLU
Time for training 10 days 1 day 4 days 5 days 15 h
Time for test 10 h 1 h 1 h 5 h 4 h