Shopee Product Detection

8th place solution with a private score of 0.84720 on Shopee Product Detection Classification 2020

Setup

Python 3.7+

To install a virtual environment with pipenv:

$ pipenv install -r requirements.txt --python 3.7

Also, we use a custom lr scheduler from this github:

$ pip install git+https://github.com/ildoonet/pytorch-gradual-warmup-lr.git

Optional: To install apex, please check this out

Training

Before training/reproducing a model, you might need to customise the arguments in run.sh to fit with your environment.

• --fp16: Set it as True if you want to train a model with mixed-precision. This requires apex to be installed
• --data-path: a path where it contains train/test.csv and image-folders
• --image-train-dir: training images folder
• --image-test-dir: test images folder
• --output-dir: output dir for saving models

To train a model:

$ bash /path/to/project/run.sh

Submission

Code for submission has been written on Kaggle Kernel