PLEASE READ MY BLOG POST : https://medium.com/@bouteillon1981/10-lessons-learned-from-participating-to-google-ai-challenge-268b4aa87efa
This repository contains samples code I used for the "Quick, Draw! Doodle Recognition Challenge" hosted by Kaggle, sponsored by Google AI.
Challenge ended on December 4th, 2018. My team and I end up in 46th position in this competition. I really want to thank all my teammates, it won't have been possible without them!
Here is our team for this challenge:
- ebouteillon (myself, charmed team leader)
- phun (team's blending alchemist)
- kalili (team's deep model wizard)
- YIANG (team's machine learning magician)
This repository contains only a portion of the code I wrote for this competition, my teammates are absolutely not responsible for my awful and buggy code.
Portion of this code is inspired by great Kagglers sharing their insight and code in Kaggle kernels and forums. I hope I did not forget to mention one of them where credit are due.
## Content of this repository
It contains three Jupyter Notebooks:
- 1-concat-csvs-into-sqlite: prepare dataset and convert it into a single sqlite database.
- 2-training-resnet18-from-scratch-with-128px-images: Train a resnet18 model using fastai library on the full dataset with square pictures of size 128px.
- a README, that you are reading now 😄
- CPU: i7-4790K
- GPU: rtx2080ti
- RAM: 24GB