Dataset options
1. Mood classification
- This goal of this project is to create a DL model which can detect six types of moods: angry, fear, happy, neutral, sad, and surprise.
- Based on Kaggle’s Face expression recognition dataset.
- There are around 3000 (500 * 6) images in the training set, and around 1000 images in the validation and test sets.
- Download link: face-expression-kaggle.zip
2. Political meme classification
- The aim of this project is to test a neural network’s ability to learn to recognize memes belonging to the two most prominent viewpoints in United States politics: conservative and liberal.
- Based on Kate Arendes’ UMSL deep learning course project.
- A collection of around 1000 images, evenly distributed between the two classes.
- Download link: political_memes.zip
3. Bears classification
- The goal is to classify images of polar, panda, and grizzly bears.
- Based on Shaney Flores’ UMSL deep learning course project
- Over 1000 digital colored images of panda, grizzly, and black bears were downloaded from an Internet image search.
- Download link: bears.zip
4. Yoga pose classification
- The goal is to build a DL model that can correctly classify yoga poses.
- Based on Navneet Kaur’s USML deep learning course project.
- A collection of 1149 images from an Internet image search, unbalanced across 4 classes.
- Download link: yoga.zip