Please note: Do not post your code/assignments publicly.
The goal of this assignment is to give you hands-on experience with various techniques for generating images using computer vision + machine learning (CVML) techniques… with dogs. At the end, you write an essay that connects several of the topics from across the course with the topics within this assignment.
For a variety of reasons, this assignment is more approachable using Google Colab (Google's version of Jupyter notebooks). Instructions below assume use of Google Colab. However, this assignment has been set-up so you should not need to upload any external data files, so everything should be ready to run.
If you haven't already, make sure you watch the Google Colab instructional video, linked in the Glow assignment.
The code to install and import everything you need is already in the Google Colab. However, for thoroughness (and for the adventurous who wish to set this up in their own environments), we need:
- scikit-image
!pip install scikit-image - tensorflow
!pip install tensorflow - tensorflow-hub
!pip install tensorflow-hub - opencv-python
!pip install opencv-python
We'll also need to import additional modules available in Google Colab, including: BytesIO, IPython.display, numpy, urllib, PIL, scipy.stats, scipy.misc, random, requests, tqdm, and cv2.
You can access any public GitHub Jupyter Notebook on Google Colab by replacing the github.com in the URL with colab.research.google.com/github, like so: https://colab.research.google.com/github/UberHowley/haii-a5/blob/main/Dog_Image_Generation.ipynb.
If you've watched the Google Colab instructional video, you know there's a couple hiccups to get through before you can work on the Colab. Follow those steps, and you should be ready to get going! Remember:
- Google Colab has a tendency to 'collapse' cells, make sure you uncollapse all the cells!
- Your runtime will expire when you idle, making it so you have to re-run all your cells, so try not to idle at an inconvenient place in the notebook. Much of this code can take a long time to run.
- Save often!
- You can download your Google Colab as a
.ipynbfile.
Follow the instructions within the notebook. Remember, the course disucssion forums and Student Help Hours are there to help you out!
Once you've completed all of the above, you're done with this project! Follow the submission instructions from within the notebook.
If you haven't used Google Colab before, a good first step for you would be to read Google's Welcome to Colaboratory and take the User Interface Tour in the Jupyter notebook Help menu once you've opened your first Jupyter Notebook.
- General resources:
- Google's Welcome to Colaboratory
- Python3 Documentation (tutorial and library reference are likely useful)
- Python tutorials from w3schools and Scrimba.
- ImageNet
- Tensorflow:
- bigGAN: