A simple image classifier trained on 6 classes of Genshin Impact characters, using a custom Convolutional Neural Network (CNN) architecture.
This project is a simple image classifier trained on 6 classes of Genshin Impact characters. The dataset was obtained from web scraping images from Pixiv.net using the tags of each 6 characters. The dataset consists of 9590 training images and 9590 test images, with each class having around 1600 images. The images were preprocessed resized to 128x128 pixels and normalized. The model was trained using a custom Convolutional Neural Network (CNN) architecture. The model was trained for 250 epochs with a batch size of 64. The model achieved an accuracy of 0.75 on the validation set.
As the images were obtained by web scraping by the tags of each character, the dataset may contain bias towards the images that despite being tagged with the character's name, may not actually be the character, either because the first image was not the character or the image was not tagged correctly. Also some images may contain multiple characters, which may cause the model to predict the wrong character.
TODO: Add model architecture and training details
- A GPU with CUDA support
- Python 3.11
- CUDA 12.1
- Conda
- Clone the repository
git clone https://github.com/02loveslollipop/GenshinImageClassifier- Create conda environment
conda env create -f environment.yml- Activate conda environment
conda activate genshin-image-classifier