Overview

• Pytorch and our toolkit: kuma_utils(https://github.com/analokmaus/kuma_utils)
• manage experiments using config class (configs.py)

What worked for us

• Tile based model like Iafoss's kernel (panda_models/PatchPoolModel2)
  • Tile setting is 224 x 224 x 64
  • We solve the problem as ordinal regression(Coral loss)
  • se-resnext50 is always the best backbone
  • public LB: ~0.87
• The biggest challenge is how to deal with noisy labels
  • We read and implemented lots of papers (models/noisy_loss.py)
  • Online Uncertaity Sample Mining (OUSM) worked best (models/noisy_loss/OUSMLoss)
    • Data with noisy label should give big loss value
    • Exclude k samples with biggest loss in a mini batch during training can prevent overfitting to noisy samples
    • In order to learn general features, first 5 epochs w/o OUSM
  • public LB: ~0.89
• Data augmentation has two types: slide aug and tile aug
  • Slide aug: ShiftScaleRotate
  • Tile aug: ShiftScaleRotate, Flip, Dropout
  • public LB: ~0.90
• Ensemble noisy detection
  • LB score is very unstable depending on the seed value
  • We trained our previous SOTA setting with 10 different seeds, and detected noise labels based on loss
    • Noise flags are added to train.csv
  • Noise ratio 0.10 + OUSM(k=1) worked the best
  • publicLB: ~0.91

Instructions

• fp16 training on nvidia apex is recommended (https://github.com/NVIDIA/apex)
• Configure INPUT_PATH in configs.py for your own environment
• Make sure all requirements in requirements.txt met
• Train with config PatchBinClassification

python3 train.py --config PatchBinClassification --fp16

Results(checkpoint, oof) will be in results/(config name)/