Smiley Predictor – interactive Python app to draw a smiley 🙂 or sad 🙁 face on a 100×100 grid and get real-time predictions using a PyTorch CNN trained on synthetic varied data.

Works even if the drawing is slightly off-center or noisy.

• 🎨 Draw smiley 🙂 or sad 🙁 faces on a 100×100 canvas
• ⚡ Real-time predictions
• 🧹 Clear canvas (Right-click or Clear button)
• 🖌 Adjustable brush size & color
• ↩️ Undo / Redo (advanced GUI)
• 🚀 GPU acceleration
• 📦 Pre-trained model: smiley_model_varied.pth

Input: 1 × 100 × 100
 └─ Conv2d(1 → 16, 3x3, padding=1) → ReLU
     └─ Conv2d(16 → 32, 3x3, padding=1) → ReLU
         └─ MaxPool2d(2x2)
             └─ MaxPool2d(2x2)
                 └─ Flatten
                     └─ Linear(32*25*25 → 64) → ReLU
                         └─ Linear(64 → 2) → Softmax
Output: Smiley / Not Smiley

This simple CNN handles shifted, noisy, or varied faces.

✅ Reached ~100% accuracy on synthetic varied data.

Drawing	Prediction
	Smiley 🙂
	Not Smiley 🙁

git clone https://github.com/IbrokhimN/Smiley.git
cd Smiley
pip install -r requirements.txt

Python 3.10+ recommended

python3 draw_smiley_varied.py

• Draw face using left mouse button
• Prediction updates in real-time
• Clear canvas: Right-click or Clear button

• Model trained on synthetic images with:

  • Random positions
  • Small noise
  • Varied face placement

• Works best with rough smiley/sad shapes

• For better accuracy:

  • Increase dataset size
  • Include different drawing styles

python3 draw_smiley_varied.py

Draw a face and watch real-time predictions!

MIT License – free to use, modify, and share