https://github.com/spandan-madan/end_to_end_image_classification_pipeline

In-depth look at an image classification pipeline for your own dataset. https://goo.gl/BuL267
https://github.com/spandan-madan/end_to_end_image_classification_pipeline

Last synced: 7 months ago
JSON representation

In-depth look at an image classification pipeline for your own dataset. https://goo.gl/BuL267

• Host: GitHub
• URL: https://github.com/spandan-madan/end_to_end_image_classification_pipeline
• Owner: Spandan-Madan
• Created: 2017-10-02T02:31:47.000Z (about 8 years ago)
• Default Branch: master
• Last Pushed: 2017-11-24T21:32:41.000Z (almost 8 years ago)
• Last Synced: 2025-01-26T12:41:24.984Z (9 months ago)
• Language: HTML
• Homepage:
• Size: 284 KB
• Stars: 7
• Watchers: 2
• Forks: 4
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
![mit-logo](https://csailprettycommittee.mit.edu/sites/default/files/images/MIT_logo.png)

# End to end image classification pipeline

Most online tutorials on image classification spend time talking about how to train a network. With the advent of PyTorch, Tensorflow and other frameworks for deep learning, training a network is in fact the easiest part of the pipeline. If you're looking for something along those lines, I have written tutorials in [Keras](https://github.com/Spandan-Madan/DeepLearningProject) and in [PyTorch](https://github.com/Spandan-Madan/Pytorch_fine_tuning_Tutorial) that you can refer to.

There are many other steps that you need to understand and get right if you want to be sure you're doing image classification right. Otherwise, chances are high you'll fall prey to one of the biggest problems in Machine Learning - "Garbage In, Garbage Out".

Some of the things this tutorial will teach you include - 

- [] Preparing your data the right way. Data augmentation using transforms, loading and making sure you exactly know what you're feeding in to your network to learn. Remember, if your input doesn't look meaningful, the network really can't learn anything meaningful!

- [] Visualizing your results

- [] Evaluating your results. Classification accuracy is a single number which tells you hardly anything about how your network is working. In order to improve your network's performance it is important you understand how it works, when it fails, what it overfits to, what it fails to learn. And so on and so forth.