CIFAR-10 CNN (1.2M parameters)

A compact VGG-style convolutional network trained from scratch on CIFAR-10. It reaches 92.23% top-1 test accuracy with under 1.2M parameters and trains in about 8.5 minutes on a single T4 GPU.

Model Details

  • Author: Vishwajeet Adkine
  • Type: convolutional image classifier (PyTorch), trained from scratch
  • Parameters: 1,197,578
  • Input / output: 3Γ—32Γ—32 RGB image, normalised with CIFAR-10 mean and std β†’ logits for 10 classes (airplane, automobile, bird, cat, deer, dog, frog, horse, ship, truck)
  • Architecture: three blocks of two 3Γ—3 conv layers (64, 128, 256 channels) with BatchNorm and ReLU, max-pooling between blocks, global average pooling, then Linear(256β†’192) β†’ ReLU β†’ Dropout(0.3) β†’ Linear(192β†’10)

Intended Use

Educational and portfolio use: a small, reproducible baseline for CIFAR-10 and for experiments such as ablations, augmentation or compression.

Not intended for real-world recognition tasks. It only knows 10 CIFAR classes at 32Γ—32 resolution.

Training

Data CIFAR-10: 45,000 train / 5,000 validation / 10,000 test
Augmentation random crop (padding 4), horizontal flip
Optimiser SGD, Nesterov momentum 0.9, weight decay 5e-4
Schedule one-cycle LR, max 0.1, 20% warm-up
Regularisation label smoothing 0.1, dropout 0.3
Epochs / batch 30 / 128, mixed precision
Selection best validation accuracy
Hardware Tesla T4, about 8.5 minutes

Results

Metric Value
Test accuracy (top-1) 92.23%
Test accuracy (top-3) 98.77%
Best validation accuracy 91.66%

The test set was used once, after model selection on the validation split.

Limitations

  • Trained only on 32Γ—32 CIFAR-10 images, so accuracy will drop on photos from other sources, other resolutions or objects outside the 10 classes.
  • No robustness testing (noise, blur, adversarial examples, distribution shift).
  • Cats and dogs are typically the hardest classes on CIFAR-10; check the confusion matrix before relying on per-class results.
  • The network always outputs one of the 10 classes, even for unrelated images.

How to Use

The class definition is in the training notebook (cnn_cifar10_classifier.ipynb).

ckpt = torch.load("cifar_cnn.pt", map_location="cpu")
model = CifarCNN(); model.load_state_dict(ckpt["state_dict"]); model.eval()

Data and Licence

CIFAR-10 (Krizhevsky, Nair and Hinton) β€” see the dataset page for terms. Code and weights: MIT.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support

Dataset used to train vishwajeet456/Cnn-classifierupdate2