# iitb-research-code/indic-trocr architecture

> Architecture diagram of the GitHub repository iitb-research-code/indic-trocr, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2025-10-15). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/iitb-research-code/indic-trocr
- Repository: https://github.com/iitb-research-code/indic-trocr

## Overview

##### Project type and purpose
- This is an OCR research/training pipeline (no frontend or service layer).
- Purpose: train and evaluate a Transformer-based OCR model (TrOCR) specialized for handwritten Indian languages using image encoder + autoregressive text decoder.

##### File structure
- Top-level:
- requirements.txt – lists Python deps (Hugging Face Transformers, PyTorch, etc.)
- scripts/
– train.py – main training pipeline
– test.py – evaluation/inference script
- No separate “src” or “backend” directory: training logic and model code live inside scripts/.

##### README for additional insights
- Describes dataset inputs (IIIT-HW word-level images + label files)
- Explains external dependency: a language-specific RoBERTa decoder model from Hugging Face
- Shows CLI usage to train or test

##### Elements
- Clear component labels (e.g., “Preprocessing Module”, “Vision Encoder”)
- External services (orange clouds)

##### Notes on detail
- Critical files and modules (requirements.txt → environment; train.py/test.py → pipelines)
- Checkpointing and logging steps
- Where hyperparameters are configured (training_args in train.py)

## Diagram (Mermaid)

```mermaid
flowchart TD
    %% Environment Layer
    subgraph "Environment"
        REQ["requirements.txt"]:::data
        RD["README.md"]:::data
    end

    %% External Services
    subgraph "External Services"
        HUG["Hugging Face Model Hub"]:::external
    end

    %% Data Layer
    DataStore["Data Store\n(train/val/test images & labels)"]:::data

    %% Processing Modules
    Preproc["Preprocessing Module"]:::compute
    DataLoader["DataLoader"]:::compute

    %% Model and Training
    Model["Model Module\n(Vision Encoder + Text Decoder)"]:::compute
    Train["Training Engine\n(data loading, training loop,\ncheckpointing)"]:::compute
    Checkpoint["Checkpoints on Disk"]:::data

    %% Evaluation Flow
    Eval["Evaluation Engine\n(load checkpoints, inference,\nmetrics)"]:::compute
    Metrics["Metrics Outputs"]:::data

    %% Flow Connections
    DataStore -->|raw data| Preproc
    Preproc -->|processed data| DataLoader
    DataLoader -->|batches| Train
    Train -->|save model| Checkpoint
    Train -->|forward/backward| Model
    Model -->|use pretrained weights| HUG
    Checkpoint -->|load checkpoint| Eval
    Eval -->|inference| Model
    Eval -->|results| Metrics

    %% Environment Usage
    REQ -->|provides deps| Preproc
    RD -->|CLI usage| Train

    %% Click Events
    click Preproc "https://github.com/iitb-research-code/indic-trocr/blob/main/scripts/train.py"
    click Model "https://github.com/iitb-research-code/indic-trocr/blob/main/scripts/train.py"
    click Train "https://github.com/iitb-research-code/indic-trocr/blob/main/scripts/train.py"
    click Eval "https://github.com/iitb-research-code/indic-trocr/blob/main/scripts/test.py"
    click REQ "https://github.com/iitb-research-code/indic-trocr/blob/main/requirements.txt"
    click RD "https://github.com/iitb-research-code/indic-trocr/blob/main/README.md"

    %% Styles
    classDef data fill:#BBDEFB,stroke:#0D47A1,color:#0D47A1;
    classDef compute fill:#C8E6C9,stroke:#1B5E20,color:#1B5E20;
    classDef external fill:#FFE0B2,stroke:#E65100,color:#E65100;
```
