vummidiganesh55/multimodal-dca-ai

How vummidiganesh55/multimodal-dca-ai works

Multimodal DCA AI turns invoice and other document pages into structured, validated results, with anomaly signals and evaluation. The README documents the principal processing stages; sampled source confirms OCR, layout, extraction, table-building, validation, and anomaly behavior. The pipeline excerpt omits key orchestration sections, so uncertain component wiring is not shown.

Main entry points: pipeline.py Document pipeline · evaluator.py Evaluation · paddle_ocr.py Paddle OCR · feature_extractor.py Anomaly features

Components 22

Input and Processing

Document pipeline src/pipeline.py
Document loading src/document_loader.py
Image preprocessing src/preprocessing/image_processor.py
PDF processing src/pdf_processor.py

Document Understanding

Paddle OCR src/ocr/paddle_ocr.py
Handwriting recognition src/handwriting/handwriting_recognizer.py
Multilingual OCR src/ocr/multilingual_ocr.py
Language detection src/multilingual/language_detector.py
Layout analysis src/layout/layout_analyzer.py
Information extraction src/extraction/information_extractor.py
Table extraction src/tables/table_extractor.py
Table JSON building src/tables/table_json_builder.py
Chart analysis src/chart_analysis/chart_analyzer.py

Validation and Risk

Document validation src/validation/document_validator.py
Anomaly features src/feature_extractor.py
ML anomaly detection src/anomaly/ml_anomaly.py
Rule-based signals src/anomaly/rules.py

Results and Evaluation

Structured results src/pipeline.py
Results dashboard app/streamlit_app.py
Evaluation src/evaluation/evaluator.py
Ground truth data/ground_truth.json

Other components

Document user

Connections 19

Document user → Document pipeline: submits document
Document pipeline → Document loading: loads document
Document pipeline → PDF processing: processes pages
Document pipeline → Image preprocessing: preprocesses images
Document pipeline → Paddle OCR: runs OCR
Paddle OCR → Layout analysis: provides text and boxes
Document pipeline → Layout analysis: analyzes layout
Document pipeline → Information extraction: extracts fields
Document pipeline → Table extraction: extracts tables
Document pipeline → Table JSON building: builds table JSON
Document pipeline → Document validation: validates data
Document pipeline → Anomaly features: prepares anomaly features
Anomaly features → ML anomaly detection: supplies features
Document pipeline → ML anomaly detection: requests anomaly analysis
Document pipeline → Rule-based signals: checks rule signals
Document pipeline → Structured results: assembles results
Evaluation → Ground truth: compares against
Results dashboard → Structured results: displays results
Evaluation → Structured results: evaluates results

More Python diagrams: ai-to-ai/auto-gmail-creator · cloud-in-a-bottle/cloud-in-a-bottle · project-monai/research-contributions · sakanaai/self-adaptive-llms · charles-001/dolphie

Drawn by GitDiagram on September 29, 2026 from the repository's file tree, README and source files. Read as Markdown ·