First Programming: Python for AI – Step 2 (Exercise Edition)

(1) Overview of the course

In this workshop, an image classification competition using Kaggle, a machine learning competition site, will be held. The objective of this workshop is to cover basic knowledge of image classification, and then actually write your own image classification code. After writing the code, participants will submit their code to Kaggle, see the results automatically scored, and improve the accuracy of the code. Note that this workshop will consist of two sessions, both of which will be conducted in person. If it is difficult to participate in person, online participation will be accepted, but in principle, it will be conducted in person.

(2) Content of each session

Basically, each session will provide an opportunity to talk in groups.

Session 1: “Image Processing 1 10/18

One image processing problem using PyTorch will be presented on Kaggle. The first step is to run the sample code distributed as a tutorial on Kaggle, check that it works correctly, and submit the code to Kaggle. After that, the goal is to create your own CNN-based model. Students will be asked to repeat the cycle of improving the code in their notebooks and asking questions if they have any doubts.

Session 2: “Image Processing 2.” 10/25

This session deals with the same problems as the previous one. In the second session, the goal is to build a model using the pre-trained model and to add preprocessing to the image to improve generalization performance. Students will be asked to repeat the cycle of improving the code on the notebook and asking questions if they have any doubts.