Image Classification: Real Art vs. AI-Generated Art

🖼️ Image Classification: Real Art vs. AI-Generated Art

In this problem, you must implement an image classification model capable of distinguishing between real artwork and AI-generated artwork.
The dataset is organized into a directory containing images and CSV files, and the model’s performance will be evaluated using the F1-score.

---

🔹 Dataset

The dataset contains two types of images:

• RealArt – real artwork (label 0)
• AIArt – images generated by AI models (label 1)

These images are stored in the images directory.

Associated files:

• train.csv – used for training
• test.csv – used for predictions

---

📁 CSV File Structure

train.csv

Columns:

• SampleID – unique numeric identifier
• ImagePath – path to the image (e.g., images/img_000123.jpg)
• Label – 0 for real art, 1 for AI-generated art

---

test.csv

Columns:

• SampleID
• ImagePath

No labels — these must be predicted by the model.

---

🎯 Problem Objective

The model must learn to classify each image into one of the following classes:

• 0 → Real Art
• 1 → AI-Generated Art

---

⚙️ Evaluation Criteria

Evaluation is performed using the F1-score computed on the predictions submitted for the test set.

Formula:

---

📨 Submission Format

The submission CSV file must contain exactly two columns:

Column	Type	Description
SampleID	integer	Test image ID
Label	integer	Prediction (0 or 1)

🔹 Example

SampleID	Label
101	0
102	1
103	1

---

📝 Note

• F1-score ≥ 0.85 → maximum score
• F1-score < 0.5 → 0 points
• Intermediate values are scaled proportionally

---

🌐 Data Source

This problem uses a public dataset available on Kaggle.