Computer Vision

NVIDIA NIM OCR

Minimal OCR dashboard powered by NVIDIA NIM service with GPU-accelerated text recognition

GPU-Accelerated OCR

A minimal single-page dashboard for sending images to the local NVIDIA NIM OCR service and visualizing recognition results. Built on NVIDIA's NemoRetriever OCR model running in Docker with GPU access.

Upload or drag-and-drop images for instant text recognition with automatic queue processing and real-time status updates.

System Architecture

Model NemoRetriever OCR v1
Container NVIDIA NIM Docker
API REST (localhost:8000)
Frontend Node.js + Static Server

Core Features

GPU-Accelerated Recognition

NVIDIA NIM service leverages GPU hardware for high-performance optical character recognition with CUDA acceleration.

Drag & Drop Upload

Simple interface for uploading multiple images via drag-and-drop or traditional file selection.

Automatic Queue Processing

Queue runs automatically by default with toggle control in sidebar. Each image card shows real-time status updates.

Base64 Image Encoding

Images are automatically converted to base64 format and sent via POST requests to NIM inference endpoint.

Real-Time Visualization

Recognized text displays in real-time with status cards. Remove individual items or clear entire queue as needed.

Docker Container Deployment

Runs as containerized service with NVIDIA GPU access, NGC authentication, and configurable cache directory.

Technical Specifications

OCR Model
NemoRetriever OCR v1
Container Platform
NVIDIA NIM (Docker)
GPU Requirements
CUDA-compatible GPU
Frontend Server
Node.js (Port 3005)
NIM Service
localhost:8000
API Endpoints
/v1/infer, /v1/ocr
Input Format
Base64 Image URL
Authentication
NVIDIA NGC API Key

Ready for GPU-Accelerated OCR?

Deploy NVIDIA NIM OCR for high-performance text recognition