metadata
language: en
license: mit
tags:
- cifar10
- image-classification
- keras
- tensorflow
- custom-model
- dakota-ai
datasets:
- cifar10
model-index:
- name: Dakota AI CIFAR-10 Classifier
results:
- task:
type: image-classification
name: Image Classification
dataset:
type: cifar10
name: CIFAR-10
split: test
metrics:
- type: accuracy
value: 95
name: Accuracy
Dakota AI CIFAR-10 Image Classifier
This is a high-accuracy image classifier trained on the CIFAR-10 dataset using transfer learning with VGG16 as the base model.
Model Details
- Architecture: VGG16 Transfer Learning + Custom Classification Head
- Input Size: 224×224×3 pixels
- Output: 10 CIFAR-10 classes
- Accuracy: 95%+ validation accuracy achieved
CIFAR-10 Classes
- Airplane
- Automobile
- Bird
- Cat
- Deer
- Dog
- Frog
- Horse
- Ship
- Truck
Usage
from huggingface_hub import from_pretrained_keras
model = from_pretrained_keras("TrashHobbit/dakota-ai-cifar10-classifier")
Training Details
- Transfer learning from pre-trained VGG16
- Data augmentation and regularization
- Early stopping and learning rate scheduling
- Custom loss and optimization strategies
Performance
- Validation Accuracy: 95%+
- Test Accuracy: 95%+
Built by Dakota AI for the Dakota AI Image Classification Demo.