|
|
--- |
|
|
license: agpl-3.0 |
|
|
language: |
|
|
- en |
|
|
metrics: |
|
|
- accuracy |
|
|
library_name: transformers |
|
|
tags: |
|
|
- medical |
|
|
pipeline_tag: image-classification |
|
|
--- |
|
|
|
|
|
This model is part of a school project. |
|
|
Utilizing the google/vit-base-patch16-224 vision transformer for image classification, |
|
|
this pre-trained model is further tuned utilizing images of pills and tablets. |
|
|
As pills and tablets have three main features, color, shape, and imprint, |
|
|
the model aims to identify images of pill and tablets by automatically extracting features. |
|
|
The dataset utilized is from the U.S. Department of Health's Computational Photography Project for Pill Identification (C3PI). |
|
|
|
|
|
DISCLAIMER: The accuracy of this model is currently low (<20%). Further training is currently ongoing to improve the accuracy. |
|
|
|
|
|
Version 2: This version tries to train the pretrained model with only 20 of the most common pills. |
|
|
Unfortunately, the accuract of the model is still currently low (<30%). |
|
|
|
|
|
|
