---
title: README
emoji: 💻
colorFrom: gray
colorTo: blue
sdk: gradio
pinned: false
sdk_version: 4.44.0
---
<p style="text-align:center;">


Our project aims to develop an image classification system capable of distinguishing between paintings created by humans and those generated by artificial intelligence.

By leveraging a combination of classification techniques and machine learning, we aim to create a model that can accurately classify different types of images and detect the critical differences between works of art.
For this project, we utilized several models, including **CNN, ELA, RESNET50, and VIT**.

---
After building and running these models and evaluating their prediction results, this is the evaluation of Results:

<span style="font-size:18px"> It can be observed that, according to the *Accuracy* metric, two models meet the desired threshold of at least 85%, which are: the *CNN+ELA* model (85%) and the *ViT* model (92%).</span>

<span style="font-size:18px">According to the *Recall* metric, we set a performance threshold of at least 80%, and there are two models that meet this requirement: the *CNN+ELA* model (83.5%) and the *ViT* model (95.7%).</span>


The following table presents the test metric results for all the models implemented in this project.

<img src="https://cdn-uploads.huggingface.co/production/uploads/66d6f28a19214d743ca1eb43/q6g7SAHT-enMFOkoXqxWc.png" alt="Description" width="500" style="display: block; margin-left: auto; margin-right: auto;"/> 


<span style="font-size:20px"> **After comparing the different results,<u> it can be seen that the model with the highest performance across all metrics is the *VIT* model</u>, achieving the best results according to all the criteria we set in the initial phase.**.</span>

---