This is a compilation of jupyter notebooks written for hands-on workshops held at Stanford/SLAC during last month (February/March 2018). They are targetted toward beginners, starting from how to use tensorflow and ends (currently) with training a deep convolutional neural network for semantic segmentation (pixel-level object categorization).
All notebooks referenced here are executable on google colaboratory with a free K80 GPU. You do not need your own GPU machine or an account (well, you do need gmail account to use colaboratory).
To run jupyter notebooks below, click a hyperlink, and choose "Connected apps" => "Colaboratory" to open them.
- Start google colaboratory
- Introduction to tensorflow: linear regression
- MNIST data set for ML-101
- MNIST image classification
- Tensorflow-Slim: high level APIs for building a graph
- Batch normalization
- Saving & Restoring Trained Weights
- Semantic segmentation with shallow CNN
- Semantic segmentation with deep CNN
Your feedbacks are always welcome! You know better how to improve them so that we can guide others better. Suggestions to include more items (like other network tasks) are also welcome! We will try our best to keep this list interesting to new members.