Apppublic

ek-developers/ocr-frontend

sourceHugging Faceupdated 7mo agoView on Hugging Face

0likes

App README

🖼️ DeepSeek OCR — Frontend

A clean, professional Gradio frontend for the DeepSeek OCR API. Upload any image, write a custom prompt, and extract text in seconds — no setup required.

This Space is a frontend only. All OCR processing is handled by the backend API Space powered by DeepSeek-OCR via Ollama.

🔗 Related Spaces

Space	Role	URL
This Space	Gradio UI frontend	`ek-developers/ocr-frontend`
Backend API	FastAPI + DeepSeek-OCR	`ek-developers/my-ocr-api`

✨ Features

—Image upload — drag and drop, click to browse, or paste from clipboard
—Custom prompts — guide the model with natural language instructions
—Auto-run on upload — OCR triggers automatically when an image is dropped in
—Copy to clipboard — one-click copy of extracted text with confirmation feedback
—Example prompts — pre-built prompts for common use cases
—Live API status — shows backend connection and endpoint URL
—Responsive layout — two-column design with input on the left, output on the right

🖥️ Interface Overview

┌─────────────────────────────────────────────────────┐
│              DeepSeek OCR                           │
│      Powered by DeepSeek · Ollama                   │
├────────────────────┬────────────────────────────────┤
│  01 · INPUT        │  02 · EXTRACTED TEXT           │
│                    │                                 │
│  [ Image Upload ]  │  ● API CONNECTED               │
│                    │                                 │
│  [ Prompt Input ]  │  [ Result Textbox         ]    │
│                    │                                 │
│  [▶ RUN] [✕ CLEAR] │  [ ⎘ COPY TEXT ]               │
├────────────────────┴────────────────────────────────┤
│  Example Prompts                                    │
└─────────────────────────────────────────────────────┘

🚀 Usage

Step 1 — Upload an image

Drag and drop an image onto the upload panel, click to browse your files, or paste directly from clipboard. Supported formats: JPG, PNG, WEBP, and other common image types.

Step 2 — Set your prompt

The default prompt is Free OCR. which extracts all visible text. You can customize it to get more targeted results:

Prompt	What it does
`Free OCR.`	Extract all visible text
`Extract all menu items and prices.`	Structured menu extraction
`Extract only the grand total.`	Receipt total only
`List all ingredients on the label.`	Product label parsing
`Read the text on this sign.`	Signage and banners
`Extract the table contents.`	Tabular data

Step 3 — Run OCR

Click ▶ RUN OCR or simply upload an image — OCR runs automatically on upload.

Step 4 — Copy the result

Click ⎘ COPY TEXT to copy the extracted text to your clipboard. A green confirmation message appears for 2 seconds.

⚙️ Configuration

Changing the backend API URL

By default the frontend points to:

https://ek-developers-my-ocr-api.hf.space/ocr

To point it at a different backend, set the OCR_API_URL environment variable in your Space settings:

1.Go to your Space → Settings → Variables and secrets
2.Add a new variable:
3.Name: OCR_API_URL
4.Value: https://your-backend-space.hf.space/ocr

📁 Project Structure

├── app.py           # Gradio UI and API call logic
├── requirements.txt # Python dependencies
└── README.md        # Documentation

`app.py` — key sections

Section	Description
`API_URL`	Backend endpoint, overridable via env var
`run_ocr()`	Sends image + prompt to the FastAPI backend, handles errors
`css`	Custom styling — IBM Plex fonts, blue accent palette
`gr.Blocks(...)`	UI layout — two-column input/output panels
`copy_btn.click(...)`	JavaScript clipboard handler
`image_input.upload(...)`	Auto-triggers OCR when image is uploaded

🛠️ Local Development

Prerequisites

—Python 3.10+
—The backend API running (locally or on HF)

Setup

bash

# Clone the Space
git clone https://huggingface.co/spaces/YOUR_USERNAME/ocr-frontend
cd ocr-frontend

# Install dependencies
pip install -r requirements.txt

# Optional: point at a local backend
export OCR_API_URL=http://localhost:7860/ocr

# Run
python app.py

The app will be available at http://localhost:7860.

📦 Dependencies

gradio>=6.6.0
requests

⚠️ Notes

Cold starts — the backend Space sleeps after ~15 minutes of inactivity on the free tier. The first request after inactivity may take 1–2 minutes while the model reloads. The frontend will show a timeout message — simply wait 30 seconds and try again.

Image quality — OCR accuracy depends on image clarity. For best results:

—Use high-resolution images (300 DPI or above for documents)
—Ensure text is not skewed or heavily shadowed
—Avoid images with very low contrast between text and background

Clipboard permissions — the copy button uses the browser's native clipboard API. If it does not work, ensure your browser has clipboard permissions enabled for the page.

📄 License

MIT License — free to use, modify, and deploy.