provides problem-solving receipts, compiled into Journeys: architecture diagrams, one-click deployment git repos, and pointers to essential docs - all are collected in a single place. In this educational challenge we are going to play with the journey "Image recognition training with PowerAI notebooks - Use transfer learning to create your own image classifier with a Jupyter Notebook
". Your goal is to follow that tutorial to solve a different problem, outlined below. Sure, as usual, you should follow any applicable best practices in your code / documentation / etc.
When you apply for some documents, you are asked to provide a valid and recent passport-style photo. In XXI century, with all phones and digital cameras around, you don't have to go to a photo master, you can just take the photo yourself, probably with a help of a friend. However, you still should follow some guildelines to ensure that your photo is valid for the document; e.g. here are the photo requirements for applicants for US visas
. Thus, your goal in this challenge is to train the image recognition algorithm to be able to take a user-provided photo as input, and tell whether this photo fits the photo requirements for US visa applications.
We will judge this challenge using a subjective scorecard. Reviewers should weight 50/50 in their scoring the technical aspect of the submission, and the actual performance of the model. As we are not sure, how well the problem can be solved with this approach and tooling, there is not need for a perfect performance to get the top score for the actual performance; for that part of scoring submissions should be compared one against another: reviewers a free to try each submission on some set of test images, and assign that part of scoring according to the number of correct / incorrect algorithm responses.
TOPCODER COGNITIVE COMMUNITY
If you have not already, go to http://cognitive.topcoder.com
and join Topcoder Cognitive Community. In case your IBM Bluemix trial has expired, ask for trial extention promo code in the challenge forum.