This repository contains all the code, configuration files, and documentation for the Invoice Annotation Project. The objective is to build an end-to-end workflow for processing a sample invoice using OCR, refining annotations with Label Studio, and preparing a dataset for AI training.

• Overview
• Project Workflow
• Requirements
• Setup and Installation
• Step-by-Step Instructions
  • Step 1: Document Analysis & Label Identification
  • Step 2: XML Label Configuration
  • Step 3: Generate OCR Data
  • Step 4: Label Refinement in Label Studio
  • Step 5: Reflection and Final Report
• License
• Contact

The project demonstrates:

• Extraction of text from a scanned invoice using Tesseract OCR.
• Creation of a custom XML configuration for Label Studio.
• Refinement of OCR output and manual labeling of invoice fields.
• Compilation of a final annotated dataset and a reflection on the process.

1. Document Analysis & Label Identification: Identify key invoice fields.
2. XML Label Configuration: Create an XML file to define labels for annotation.
3. Generate OCR Data: Use a Colab Notebook to extract text and bounding boxes.
4. Label Refinement in Label Studio: Import and refine annotations.
5. Reflection and Final Report: Write a reflection on your process and improvements.

• Python 3.7+
• Tesseract OCR (installed and in your PATH)
• Google Colab account (for running the OCR notebook)
• Label Studio (installed locally)

1. Clone the Repository:

   git clone https://github.com/yourusername/Invoice-Annotation-Project.git
   cd Invoice-Annotation-Project
   

2. Install Label Studio:

pip install label-studio
label-studio start

3. Set Up Tesseract OCR: Follow instructions for your OS to install Tesseract OCR.

Task:
Review the sample invoice image and identify key fields.

Deliverable:
A bullet list of fields saved in document_analysis.txt.

Example:

• Invoice Number
• Invoice Date
• Customer Name
• Item Descriptions
• Quantity, Unit Price
• Tax
• Total Amount

Task:
Create an XML configuration file for Label Studio to define the labels for annotation.

Deliverable:
Save the following as invoice_label_config.xml.

Example:

<View>
  <Image name="image" value="$image" zoom="true" />
  <RectangleLabels name="label" toName="image">
    <Label value="Invoice Number" background="blue" />
    <Label value="Invoice Date" background="green" />
    <Label value="Customer Name" background="yellow" />
    <Label value="Item Descriptions" background="orange" />
    <Label value="Quantity, Unit Price" background="red" />
    <Label value="Tax" background="pink" />
    <Label value="Total Amount" background="purple" />
  </RectangleLabels>
  <TextArea name="transcription" toName="image" editable="true" perRegion="true"
            rows="3" placeholder="Enter recognized text here" />
</View>

Task:
Run the provided Google Colab Notebook to extract text and bounding boxes from the sample invoice.

Instructions:

1. Open the Pre-Configured Colab Notebook.
2. Update the image_url variable to:

   https://allies-assets.s3.us-east-1.amazonaws.com/birthplan_builder_assets/extra_images/invoice_sample.png

3. Click Runtime > Run all to execute all cells.
4. When prompted (or via the Files sidebar), download the generated file as invoice_ocr_output.json.

• Deliverable: invoice_ocr_output.json

Task: Import the OCR JSON file into Label Studio, refine the bounding boxes, correct OCR text, and assign the correct labels.

Instructions:

1. Launch Label Studio at http://localhost:8080 and create a new project named "Invoice Annotation."
2. In the project settings, paste the XML configuration from Step 2 and click Save.
3. Import the invoice_ocr_output.json file into your project.
4. Open the task, adjust bounding boxes, delete any irrelevant annotations, and manually correct any OCR errors.
5. Once satisfied, export the final labeled dataset as label_studio_output.json.

• Deliverable: label_studio_output.json

Task: Write a reflection (150–200 words) discussing:

1. Label Selection: Why you chose the specific labels.
2. Challenges: What issues you encountered during OCR and annotation, and how you addressed them.
3. Workflow Improvements: Suggestions for streamlining the process in the future.

• Deliverable: Save your reflection in reflection.md.

This project is licensed under the MIT License.

For any questions, please contact [omoiza@ttu.edu].