CIFAR-10 Simple CNN Classifier (PyTorch)
A simple Convolutional Neural Network (CNN) trained from scratch on the CIFAR-10 dataset (10 classes, $32 imes 32$ pixels).
📊 Model Performance
| Metric | Value |
|---|---|
| Test Set Accuracy | 81.46% |
🛠️ Architecture Details
- Model: Custom SimpleCNN (3 convolutional layers)
- Dataset: CIFAR-10
- Library: PyTorch
- Training Details: 25 Epochs using Data Augmentation (Random Crop, Random Flip)
- Final Model Path:
pytorch_model.bin
📌 How to Use
To load the weights of this model in PyTorch:
from huggingface_hub import hf_hub_download
import torch
from your_model_definition import SimpleCNN # Replace with your actual model class
# Download the model weights
model_path = hf_hub_download(repo_id="manasyesuarthana/cifar10-simple-cnn-pytorch", filename="pytorch_model.bin")
# Initialize your model architecture
model = SimpleCNN()
# Load the downloaded weights
model.load_state_dict(torch.load(model_path, map_location=torch.device('cpu')))
model.eval()