README.md

metadata

datasets:
  - cifar10
metrics:
  - accuracy
pipeline_tag: image-classification

ResNet-20

ResNet-20 image classification model for CIFAR-10. The pretrained PyTorch weights and model implementation were downloaded from chenyaofo/pytorch-cifar-models. The PyTorch model was converted to ONNX for this repository.

CIFAR-10 32 × 32

• Origin: chenyaofo/pytorch-cifar-models (cifar10_resnet20)
• Input:
  • size: [1, 3, 32, 32]
  • format: NCHW
• Output:
  • size: [1, 10] (CIFAR-10 class scores)
• ONNX:
  • opset: 11
  • Operators: 53 (7 types)
    • Add: 9
    • Constant: 1
    • Conv: 21
    • Gemm: 1
    • GlobalAveragePool: 1
    • Relu: 19
    • Reshape: 1