README.md

metadata

language: en
tags:
  - image-classification
  - computer-vision
  - pytorch
  - cnn
  - cifar10
license: mit
datasets:
  - cifar10
model-index:
  - name: SimpleCNN CIFAR-10 Classifier
    results: []

🧠 SimpleCNN CIFAR-10 Classifier

📌 A simple Convolutional Neural Network (CNN) model trained on the CIFAR-10 dataset, capable of recognizing 10 classes of common objects. The model was trained using PyTorch and is suitable for educational and prototyping purposes.

🏷️ Classes

• Airplane
• Automobile
• Bird
• Cat
• Deer
• Dog
• Frog
• Horse
• Ship
• Truck

🧰 Training Procedure

1. Built a custom CNN model with 3 convolutional layers and 2 fully connected layers.
2. Used MaxPooling after each conv layer and dropout for regularization.
3. Resized all input images to 32x32 and applied normalization: (mean=0.5, std=0.5).
4. Training/validation split:
   • 80% Training
   • 20% Validation
5. Training setup:
   • Optimizer: Adam
   • Loss Function: CrossEntropyLoss
   • Batch size: 64
   • Learning rate: 0.001
   • Epochs: 10
6. Saved the best-performing model as pytorch_model.bin.

📊 Performance

Metric	Value
Best Validation Accuracy	88.76%

⚙️ Framework & Environment

• Python: 3.11
• PyTorch: 2.x (Colab)
• Torchvision: 0.15.x
• Platform: Google Colab (GPU enabled)

🧪 Hyperparameters

Parameter	Value
Epochs	10
Batch Size	64
Optimizer	Adam
Learning Rate	0.001
Loss Function	CrossEntropy
Image Size	32x32
Data Split	80% Train / 20% Val

---