🏆 FER-ML Competition Rules

Welcome to the FER-ML Competition!
Please read the following rules carefully before participating.
These rules ensure fair play and maintain the integrity of the leaderboard.

---

🧩 Objective

Develop a machine learning model that accurately classifies human facial expressions using the FER-2013 dataset.
Your goal is to maximize classification accuracy on the hidden private test set.

---

🗂 Dataset Information

• The dataset is derived from FER-2013.
• It contains 48×48 grayscale facial images labeled with one of seven emotions:
  • 0 → Angry
  • 1 → Disgust
  • 2 → Fear
  • 3 → Happy
  • 4 → Sad
  • 5 → Surprise
  • 6 → Neutral

Public Files:

the dataset is hosted externally on Google Drive.

🔗 Download Link

➡️ Click here to access the dataset

Private Files:

• Test_data → Hidden ground-truth data used for automatic evaluation

---

🧠 Model Requirements

Participants must use machine learning models (e.g., Random Forest, SVM, Logistic Regression, etc.).
Deep learning models (CNNs, transformers, etc.) may be used only if implemented locally — no pretrained weights or external datasets allowed.

---

⚙️ Submission Guidelines

1. Train your model using data/train.csv.

2. Predict labels for the samples in data/test.csv.

3. Save your predictions as submissions/submission.csv in the format:

   id	predicted_label
   0	3
   1	0
   2	6

4. Create a Pull Request (PR) with your submission file.

5. The GitHub Actions workflow will automatically evaluate your submission using the hidden private test labels.

---

🧮 Evaluation Metric

• Metric: Accuracy
• The evaluation script compares your predicted_label values with the hidden ground truth.
• The result will appear automatically in your Pull Request Checks tab.

---

🚫 Restrictions

To ensure fairness:

• ❌ No external labeled data
• ❌ No manual tuning based on private test feedback
• ❌ No submission of multiple models in a single PR
• ❌ No use of pretrained deep learning models unless approved

---

🕐 Submission Limits

• Each Team should name the file something like: Team1.
• Only the latest submission counts toward the leaderboard.

---

🧾 Leaderboard

• Leaderboard scores are automatically updated based on accuracy.
• In case of identical scores, the earlier submission time will be used as a tiebreaker.

---

📩 Contact

If you face issues with the repository or evaluation:

• Contact me at tasneem.mselim@gmail.com

---

Good luck, and happy modeling! 🚀
Let the best model win!