argentinian_ids

Argentinian ID Validation Model

This model was trained to validate Argentinian IDs based on images of the front, back, and invalid IDs. The model uses a ResNet18 architecture and classifies images into three categories: valid_front, valid_back, and invalid.

Usage

• The model can be loaded in PyTorch using the .pth file or in ONNX format for deployment in a variety of environments.
• You can test the model with the test_model.py script.

Files

• argentinian_id_validation.py: Code used for training the model.
• test_model.py: Script to test the model on new data.
• dni_image_validation_model.pth: PyTorch model weights.
• dni_image_validation_model.onnx: ONNX version of the model.

Conversion to ONNX

To convert the PyTorch model to ONNX, the following code was used:

import torch
import torchvision.models as models

model = models.resnet18(pretrained=False)
model.fc = torch.nn.Linear(model.fc.in_features, 3)  # 3 classes
model.load_state_dict(torch.load('dni_image_validation_model.pth'))

dummy_input = torch.randn(1, 3, 224, 224)  # Input shape
torch.onnx.export(model, dummy_input, "dni_image_validation_model.onnx",
                  input_names=['input'], output_names=['output'],
                  opset_version=11)