Image Organizer

Inspiration

I don't have much experience with artificial intelligence and machine learning. I am currently taking an AI intro class, and I wanted to use some of the knowledge I learned even though I am not proficient. I was primarily inspired to participate in this Hackathon to gain some meaningful experiences. I tried to come up with an idea that can be useful in my life. I realized that I haven't clean up the files on computers at home, so I decided to make a program that I can use in real life using AI concepts.

What it does

My project basically helps the user to organize image files. There are two major features: find duplicates and classify images. Find duplicates feature takes the source and destination path and a filename, and the app puts all the images that are the same as the input file name into the destination directory. The user then can type in filenames to remove duplicate files. The second feature trains a model with user input categories. After training a model, it will predict which category an image is in. I then enter the type of object, source path, and destination path to move the images that are classified as the user input type to the destination.

How we built it

I built it by using Jupiter notebook. I first wrote small functions to test if my logic works and gradually expanded my initial ideas. I also did a lot of research on machine learning using Keras.

Challenges we ran into

The biggest challenge was to increase the accuracy of the model and decrease errors. I at first tried to gather more images for the training set, but I found that it is better to create augmented images for efficiency. Although it made the process of the model longer, it improves the accuracy of the model, which is my biggest goal.

Accomplishments that we're proud of

I am proud that I actually tried machine learning. I am glad that I actually finished my project and accomplished what I had planned to do for this hackathon.

What we learned

I learned a lot through this project. I primarily learned how to learn by reading books and search on webs. I have gotten several bits of help online, but it had been so long since I learned from actual books. I realized there are so many good textbooks I can read and learn more in-depth about my interests.

What's next for Image Organizer

The next is still to improve accuracy because there are still errors in the model even though the accuracy of the model is almost always greater than 95%. I should think about what is an effective way of creating a more accurate model without delaying the training process too much.

Built With

• jupytor
• keras
• python