https://github.com/h-fuzzy-logic/python-windspeed

Predicting hurricane windspeed from satellite images
https://github.com/h-fuzzy-logic/python-windspeed

cnn keras python tensorflow

Last synced: 27 days ago
JSON representation

Predicting hurricane windspeed from satellite images

• Host: GitHub
• URL: https://github.com/h-fuzzy-logic/python-windspeed
• Owner: h-fuzzy-logic
• Created: 2021-02-23T00:56:28.000Z (almost 4 years ago)
• Default Branch: main
• Last Pushed: 2021-02-23T01:20:37.000Z (almost 4 years ago)
• Last Synced: 2023-08-27T20:09:14.748Z (over 1 year ago)
• Topics: cnn, keras, python, tensorflow
• Language: Jupyter Notebook
• Homepage:
• Size: 206 KB
• Stars: 3
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Predicting windspeed of satellite images of hurricanes

This notebook was created for a Driven Data [competition](https://www.drivendata.org/competitions/72/predict-wind-speeds/).  

# Tools Used

* Google Collab Notebook w/ TPUs

* Python, Keras, Tensorflow

* Convolutional Neural Network (CNN)

# Technical Highlights

* Training set consisted of 70,000 366x366 JPG files (1 GB)

* To improve performance, Google Collab's TPUs were used  

* JPG files were convered to TFRecord format and the TFRecord files were stored in a Google Bucket

* The TFRecords were streamed in batches to the model during training 

* During training, model checkpoints (for the Best MSE) are saved to Google Drive

* At the end of training, the final model is also saved to Google Drive