Kaggle: imaterialist-challenge-furniture-2018

Competition: image classification with 128 classes Link: https://www.kaggle.com/c/imaterialist-challenge-furniture-2018

How to run

1. Download data from kaggle to ./data/
2. Train model python cnn_runner_pth.py train model_name augmentation_mixup mixup_alpha
3. Predict python cnn_runner_pth.py predict model_name augmentation_mixup mixup_alpha
4. Calculate weights for finetuned models python statistics.py
5. Generate submission python submit.py

Ensemble

• ensemble denset201, inceptionv4, inceptionresnetv2, senet154, se_resnext101, naslarge
• hflip at test

Result

• Public Borard: 9/436 Private Board: 13/436
• models trained at model

Related Projects

• skrypka : winner of imaterialist furniture 2018