OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Deep Text Recognition Benchmark

Deep Text Recognition Benchmark is an open-source PyTorch implementation of a four-stage framework for benchmarking deep-learning scene-text recognition methods, accompanied by training and evaluation data, pretrained models and recognition failure cases.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Deep Text Recognition Benchmark — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when you need a research codebase and shared data for scene-text recognition comparisons, or want to explore the documented pretrained demo on images.

Who it is for

  • Researchers comparing scene-text recognition model components under shared datasets.
  • Developers exploring OCR on images using the provided pretrained-model demo.

Documented uses

  • Compare scene-text recognition methods using consistent training and evaluation datasets.
  • Inspect recognition failure cases and their cleansed labels.
  • Run the documented demo on image files with a pretrained model.
Before you choose

Requirements & limits

  • The README documents Python 3.6, PyTorch 1.3.1, CUDA 10.1 and Ubuntu 16.04 as its tested environment.
  • The implementation lists LMDB, Pillow, torchvision, NLTK and natsort as requirements.
  • Training data includes MJSynth and SynthText; evaluation data includes IIIT, SVT, IC03, IC13, IC15, SVTP and CUTE.
  • Pretrained weights and dataset archives are obtained through external download links referenced by the repository.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Run the documented pretrained-image demo

Use a pretrained scene-text recognition model to process images placed in the demo folder.

Before you start

  • Python 3.6, PyTorch 1.3.1, CUDA 10.1 and Ubuntu 16.04 are the README's tested environment.
  • Install lmdb, pillow, torchvision, nltk and natsort.
  • Download a pretrained model and provide image files in demo_image/.

Steps

  1. Download a pretrained model from the repository's linked model folder.

  2. Place the images to test in demo_image/.

  3. Run demo.py with the documented TPS, ResNet, BiLSTM and Attn configuration, setting --saved_model to the downloaded weights.

  4. Add --sensitive when using a case-sensitive model.

Expected outcome

The demo produces text-recognition predictions for the supplied images.

Guide limitations

  • The guide describes the repository's documented environment and procedure; it does not guarantee compatibility with other setups.
  • The command assumes the downloaded model filename matches the --saved_model value.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
15 October 2026
Original website title
clovaai/deep-text-recognition-benchmark · Official homepage title · retained in its source language
How we review sourcesSuggest a correction