metadata
language:
- es
- en
pipeline_tag: image-classification
widget:
- src: >-
https://upserve.com/media/sites/2/Bill-from-Mezcalero-in-Washington-D.C.-photo-by-Alfredo-Solis-1-e1507226752437.jpg
example_title: receipt
- src: https://templates.invoicehome.com/invoice-template-us-neat-750px.png
example_title: invoice
InvoiceReceiptClassifier is a fine-tuned LayoutLMv2 model that classifies a document to an invoice or receipt.
Quick start: using the raw model
from transformers import (
AutoModelForSequenceClassification,
LayoutLMv2FeatureExtractor,
LayoutLMv2Tokenizer,
LayoutLMv2Processor,
)
model = AutoModelForSequenceClassification.from_pretrained("fedihch/InvoiceReceiptClassifier")
feature_extractor = LayoutLMv2FeatureExtractor()
tokenizer = LayoutLMv2Tokenizer.from_pretrained("microsoft/layoutlmv2-base-uncased")
processor = LayoutLMv2Processor(feature_extractor, tokenizer)
from PIL import Image
input_img = Image.open("*****.jpg")
w, h = input_img.size
input_img = input_img.convert("RGB").resize((int(w * 600 / h), 600))
encoded_inputs = processor(input_img, return_tensors="pt")
for k, v in encoded_inputs.items():
encoded_inputs[k] = v.to(model.device)
outputs = model(**encoded_inputs)
logits = outputs.logits
predicted_class_idx = logits.argmax(-1).item()
id2label = {0: "invoice", 1: "receipt"}
print(id2label[predicted_class_idx])