https://github.com/emrys-hong/exclusive-image-kaggle
https://github.com/emrys-hong/exclusive-image-kaggle
image-net kaggle-competition
Last synced: 3 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/emrys-hong/exclusive-image-kaggle
- Owner: Emrys-Hong
- Created: 2018-10-21T03:24:09.000Z (over 6 years ago)
- Default Branch: master
- Last Pushed: 2019-01-25T12:49:49.000Z (over 6 years ago)
- Last Synced: 2025-01-15T08:19:26.758Z (4 months ago)
- Topics: image-net, kaggle-competition
- Language: Jupyter Notebook
- Size: 5.19 MB
- Stars: 1
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: Readme.md
Awesome Lists containing this project
README
# [kaggle competition exclusive image](https://www.kaggle.com/c/inclusive-images-challenge/leaderboard)
## difficulties
the bigges difficulty lies in the dataset is so BIG(512G), we narrow down the number of classes to fit the model. listed [here](https://www.kaggle.com/jpmiller/basic-eda-with-images)
As the function is multi classification, therefore we need to change our metric to F2 loss function. and change the loss error from softmax to sigmoid.