/computer_vision_light_weight_modeling_pjt

Primary LanguageJupyter Notebook

computer_vision_light_weight_modeling_pjt

Naver connect boostcamp AI Tech pjt

computer vision light weight modeling task

  1. AutoML random model search

using optuna

efficient search using pruning

  1. tensor decomposition retraining

VBMF MUSCO

  1. computer vision training trick

hyperparameter(loss(F1, label smoothing, cross entropy), optimizer, scheduler, learning rate)

skip connection customizing

data augmentation(cutMix, MixUp, random augmentation..)

boosting training

data downsampling, upsampling to handle unbalanced class problem