Shoulder Implants X-Ray Manufacturer Classification with Vision Transformer

In this project, we explored different traditional machine learning/deep learning algorithms to classify the manufacturer of a shoulder X-ray image. The range of the manufacturer takes four values, namely Depuy, Cofield, Tornier and Zimmer.

We performed traditional machine learning algorithms Random Forest and K-Nearest Neighbor to compute the benchmark for this task. Next, we performed data augmentation technique on each image to enlarge the training dataset. We then pre-trained/trained from scratch VGG-16, ResNet-50, InceptionV3 and Vision Transformer Neural Network. The best precision score achieved is 81% based on Data Augmentation and 10 fold cross validation on ResNet-50.

For more details, please refer to our arXiv link. You could also find the GitHub repository in the paper.

References:

@article{mz2021,

title = "Shoulder Implant X-Ray Manufacturer Classification: Exploring with Vision Transformer",

author = "Zhou, Meng", "Mo, Shanglin",

year = "2021",

url = "https://arxiv.org/abs/2104.07667"

}