DocuNet_256x256_FineTuned_CLS6

DocuNet

CNN efficientNet FineTuned for classifying personal mexican documents.

Model Details

Model Description

CNN efficientNet FineTuned for classifying personal mexican documents this model was done in colaboration with NMP autoavanza for detecting document types and classifying it in three classes such as INE, TC and Invoice.

This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.

• Developed by: Juan Manuel Carballo
• Model type: CNN EfficientNet Finetune
• License: Apache Licence 2.0

Uses

Project autoavanza in colaboration with NMP

How to Get Started with the Model

It's mainly used to classify scanned documents of three types INE, TC and Invoice

Training Details

Training Data

The training data was provided by NMP from their vault.

Training Procedure

The training procedure was done using 50 epochs on the dataset that consist of ~250 samples.

Preprocessing

Augmentation techniques were applied such rotation, scale, and affine transformations.

Training Hyperparameters

• Training regime: fp32 in a macbook pro m4 using pytorch.

Evaluation

Metrics

val_acc=90.48%, val_loss=0.1532, lr=0.0000181

Results

Software

pytorch==2.7.1 torchvision=0.22.1

Model Card Contact

[email protected]