TeleOCR README

The project's README from github.com/caipeng328/TeleOCR. As one markdown file: README.md.

TeleOCR

TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents

License Hugging Face Papers with Code: SOTA on OmniDocBench v1.6

A lightweight Vision-Language Model for unified document parsing across digital and camera-captured documents.

🔥 News

📖 Introduction

TeleOCR is a lightweight (~1.2B parameters), open-source Vision-Language Model designed specifically for document understanding.

Unlike existing approaches that primarily focus on either digital documents or camera-captured documents, TeleOCR provides a unified framework for both scenarios.

TeleOCR introduces the following key techniques:

These techniques enable TeleOCR to achieve strong performance across both digital and camera-captured document benchmarks while maintaining a lightweight architecture suitable for practical deployment.

TeleOCR benchmark results

⚙️ Data Engine

TeleOCR data engine

TeleOCR adopts an automated data engine consisting of four major components:

Component Description
Multi-node Consensus Voting Generates reliable pseudo-labels from heterogeneous model predictions
Geometry-aware Data Synthesis Produces realistic camera-captured document samples
Image-to-Image Self-Verification Verifies model predictions through visual rendering
Progressive Data Cleaning Iteratively filters and refines training data

Most of the generated training data requires no manual annotation.

🧠 Progressive Training

TeleOCR training pipeline

TeleOCR is trained through a progressive four-stage training pipeline:

Stage Objective
Stage 1 Vision-Language Alignment
Stage 2 Geometry-aware Document Parsing
Stage 3 Content-Structure Decoupled Learning
Stage 4 Reinforcement Learning

This progressive strategy gradually improves the model from basic vision-language alignment to fine-grained document understanding and structured output generation.

📊 Experimental Results

TeleOCR is evaluated on OmniDocBench v1.6, Wild-OmniDocBench, PureDocBench, and the ICDAR 2026 Sci-ImageMiner Challenge.

Bold indicates the best result, while underline indicates the second-best result. It should be noted that all metrics reported in this report were evaluated using the Docker image provided by OmniDocBench v1.6. For Wild-OmniDocBench, we used the Docker image provided by OmniDocBench v1.6 for evaluation, with the metrics obtained by inspecting the Sub-v1.5 results. PureDocBench was also evaluated using the Docker image provided by OmniDocBench v1.6; however, the ground-truth JSON files were converted from the files provided by PureDocBench. In addition, for the real-world degraded tracks in Wild-OmniDocBench and PureDocBench, we recommend enabling LAYOUT_MODE="Segmentation".

Layout Visualization of Distorted Documents

To evaluate the model's ability to understand complex document deformations, we conduct a visual evaluation on the public dewarping datasets DocUNet and DIR300, with representative results shown in Figure. TeleOCR directly performs layout and content parsing on distorted documents without dewarping preprocessing or a dedicated rectification model, demonstrating robust parsing under complex geometric deformations.

Parsing evaluation on the DIR300 dataset. Parsing evaluation on the DocUNet dataset.

Dr.DocBench Challenge

模型 overall ↑ Text edit ↓ formula cdm ↑ Table teds ↑ order edit ↓
Specialized VLMs
TeleOCR 67.96 0.1903 0.02 64.97 0.398
Mineru 2.5 pro 62.26 0.3402 0.04 67.75 0.356
OvisOCR2 59.25 0.3883 0.00 61.59 0.3791
PaddleOCRvl 1.6 55.11 0.4364 0.21 51.34 0.412

OmniDocBench v1.6

OmniDocBench

Model Type Method Params Overall ↑ Text Edit ↓ Formula CDM ↑ Table TEDS ↑ Table TEDS-S ↑ Read Order Edit ↓
Specialized VLMs TeleOCR 1.2B 96.87 0.027 96.36 97.05 98.52 0.122
OvisOCR2 0.8B 96.58 0.025 97.53 94.76 97.16 0.111
PaddleOCR-VL-1.6 0.9B 96.33 0.033 97.49 94.76 97.11 0.127
MinerU2.5-Pro 1.2B 95.75 0.036 97.45 93.42 95.92 0.120
GLM-OCR 0.9B 95.22 0.044 97.18 92.83 95.39 0.133
PaddleOCR-VL-1.5 0.9B 94.87 0.038 96.69 91.67 94.37 0.130
HunyuanOCR-1.5 1B 94.74 0.033 97.49 94.76 97.11 0.127
PaddleOCR-VL 0.9B 94.11 0.040 95.70 90.65 93.74 0.135
Youtu-Parsing 2.5B 93.68 0.044 93.45 92.02 95.00 0.116
Logics-Parsing-v2 4B 93.27 0.041 95.47 88.42 91.98 0.137
FireRed-OCR 2B 93.20 0.037 95.27 88.04 91.06 0.131
MinerU2.5 1.2B 92.98 0.045 95.59 87.88 91.47 0.130
OpenDoc-0.1B 0.1B 90.64 0.049 92.93 83.88 87.45 0.140
dots.ocr 3B 90.50 0.048 89.12 87.18 90.58 0.138
DeepSeek-OCR 2 3B 90.17 0.050 91.59 83.89 87.75 0.144
HunyuanOCR 1B 89.87 0.089 87.44 91.01 93.23 0.171
Dolphin-v2 3B 89.34 0.069 90.53 84.40 87.44 0.150
OCRVerse 4B 88.44 0.063 89.14 82.44 86.27 0.163
MonkeyOCR-pro-3B 3B 88.43 0.074 88.33 84.35 88.62 0.189
General VLMs Ovis2.6-30B-A3B 30B 93.62 0.035 94.93 89.44 92.40 0.135
Gemini 3 Pro -- 92.85 0.064 95.83 89.15 92.96 0.165
Gemini 3 Flash -- 92.58 0.066 95.03 89.29 93.51 0.173
Qwen3-VL-235B 235B 89.78 0.063 92.53 83.07 86.75 0.166
GPT-5.2 -- 86.52 0.114 88.00 82.95 87.93 0.193
InternVL3.5-241B 241B 83.61 0.130 89.52 74.35 79.78 0.215

Wild-OmniDocBench

Wild-OmniDocBench

Model Type Method Params Overall ↑ Text Edit ↓ Formula CDM ↑ Table TEDS ↑ Table TEDS-S ↑ Read Order Edit ↓
Decoupled VLMs TeleOCR 1.2B 88.53 0.1173 88.26 89.05 92.14 0.2011
PaddleOCR-VL-1.6 0.9B 87.36 0.1369 88.42 85.76 90.14 0.2057
MinerU2.5-Pro 1.2B 87.33 0.1362 90.15 85.46 90.12 0.2013
GLM-OCR 0.9B 85.08 0.1514 89.09 81.31 85.90 0.2228
PaddleOCR-VL-1.5 0.9B 84.64 0.1461 86.72 81.80 86.52 0.2138
End-to-End VLMs OvisOCR2 0.8B 87.91 0.129 90.37 85.13 89.11 0.2021
dots.ocr 3B 81.84 0.1483 85.00 75.32 80.20 0.2200
HunyuanOCR-1.5 1B 77.62 0.1979 85.12 67.54 70.67 0.2750
Logics-Parsing-v2 4B 77.10 0.4029 91.40 80.19 87.16 0.2355

PureDocBench

PureDocBench

Model Type Model Clean Overall ↑ Clean Text ↓ Clean Formula ↑ Clean Table ↑ Digital Degraded Overall ↑ Digital Degraded Text ↓ Digital Degraded Formula ↑ Digital Degraded Table ↑ Real Degraded Overall ↑ Real Degraded Text ↓ Real Degraded Formula ↑ Real Degraded Table ↑
Decoupled VLM TeleOCR 86.90 0.111 81.01 91.09 77.47 0.206 72.59 80.45 70.85 0.302 65.11 77.66
DotsMOCR 76.27 0.151 66.23 77.65 73.16 0.198 64.32 74.95 61.73 0.312 54.39 61.97
MinerU2.5-Pro 75.87 0.222 65.14 84.68 71.77 0.272 61.79 80.73 62.56 0.375 52.70 72.47
YouTu-Parsing 75.02 0.230 67.34 80.74 69.66 0.270 61.44 74.49 60.29 0.360 52.20 64.69
PaddleOCR-VL-1.5 73.01 0.266 63.53 82.12 66.73 0.339 58.03 76.07 60.50 0.398 54.00 67.33
GLM-OCR 68.65 0.314 57.89 79.44 63.06 0.383 53.23 74.21 58.31 0.433 50.34 67.83
Dolphin-v2 65.90 0.342 59.80 72.12 60.24 0.393 52.20 67.86 44.92 0.553 39.98 50.04
MonkeyOCR-pro-3B 62.23 0.346 48.46 72.83 57.40 0.397 45.57 66.32 46.49 0.511 38.18 52.43
End-to-End VLM OvisOCR2 82.14 0.149 71.29 90.12 77.77 0.192 67.87 84.71 66.61 0.316 57.64 73.79
FD-RL 78.38 0.193 68.21 86.22 76.33 0.214 67.16 83.22 67.04 0.298 58.82 72.08
Logics-Parsing-v2 76.35 0.213 67.67 82.67 73.85 0.248 67.33 79.02 67.64 0.304 61.65 71.64
dots.ocr 72.01 0.248 61.37 79.51 65.95 0.307 56.67 71.86 55.68 0.403 47.70 59.63
Qianfan-OCR 57.22 0.370 49.79 58.83 50.85 0.438 44.41 51.96 45.06 0.494 39.08 45.53
General VLMs Qwen3-VL-8B 72.44 0.261 65.10 78.35 72.03 0.266 64.88 77.82 62.73 0.342 55.55 66.81
Kimi K2.6 72.32 0.303 66.93 80.30 69.95 0.322 64.69 77.31 68.02 0.335 62.44 75.14
Gemini-3.1-Pro 70.04 0.306 65.63 75.08 69.28 0.322 65.81 74.24 71.98 0.300 68.62 77.26
Qwen3.5-397B-A17B 69.12 0.233 65.26 65.40 68.34 0.244 63.91 65.53 62.70 0.287 60.70 56.12

ICDAR 2026 Sci-ImageMiner Challenge

Challenge Website

Rank Team RMS TEDS Weighted
1 TeleOCR 17.23 66.39 41.81
2 VLMinators 17.29 64.31 40.80
3 Ricoh_SRCB 16.23 61.12 38.67
4 Vassilis Sioros 14.94 55.20 35.07
5 DocMiner 12.67 53.72 33.19
6 Qwen3 VL 8B 14.08 57.86 35.97

🚀 Installation & Usage

1. Environment Installation

TeleOCR requires Python 3.10+ and a CUDA-enabled environment for GPU inference.

git clone https://github.com/caipeng328/TeleOCR.git
cd TeleOCR
conda create -n teleocr python=3.10 -y
conda activate teleocr

Install Dependencies

Install TeleOCR and its dependencies in editable mode:

pip install -e .

The -e option installs the project in editable mode, allowing modifications to the source code to take effect immediately without reinstalling the package.

2. Inference

TeleOCR provides infer.py for batch inference on document images.

Before running inference, configure:

A typical inference command is:

#!/bin/bash
set -e

IMAGE_SUB_PATH="/path/to/input/images"
RESULT_SAVE_PATH="/path/to/output/results"

python infer.py \
    --image_sub_path "${IMAGE_SUB_PATH}" \
    --result_save_path "${RESULT_SAVE_PATH}" \
    --us

TeleOCR (caipeng328/TeleOCR)

Searching Database…

343 ·GitHub
TeleOCR
What is this?TeleOCR is a lightweight (~1.2B parameter) open-source Vision-Language Model built specifically for document parsing, handling both regular digital documents and messy camera-captured photos with a single unified model.
Is this a ready-to-use app or just a model with code?It's a research VLM plus a Python inference script (infer.py) — you clone the repo, install it as a Python package, and run document parsing from the command line yourself; there's no packaged GUI app.
How do I run it on my own documents?Install the project in editable mode with pip, then run infer.py pointing it at your image folder, output folder, and model path, e.g. python infer.py --imagesubpath --resultsavepath .
What do I need installed before trying this?Python 3.10 or newer and a CUDA-enabled GPU environment, since inference is GPU-based; the editable install (-e) lets you tweak the source without reinstalling.
Any catches before relying on this for real scans?The README is almost entirely benchmark tables and install snippets with no stated license, so reuse terms aren't clear; for real-world degraded photos the authors recommend explicitly enabling a 'Segmentation' layout mode rather than trusting the default settings.
Is NaviDC-OCR the same project as TeleOCR?Yes — NaviDC-OCR was renamed to TeleOCR in September 2026, and all future model releases will ship under the TeleOCR name.
How does TeleOCR compare to MinerU 2.5 Pro and PaddleOCR-VL?On OmniDocBench v1.6 TeleOCR scores 96.87 overall versus 95.75 for MinerU2.5-Pro and 96.33 for PaddleOCR-VL-1.6, despite having only 1.2B parameters, and it also won the Dr.DocBench and ICDAR 2026 Sci-ImageMiner challenges.
Can TeleOCR run outside the original Python/CUDA setup?Yes, a community member produced a GGUF conversion with llama.cpp support, and others have reported deploying it on Huawei Ascend 910B hardware.
What does the Curvature-Guided Douglas-Peucker Sampling (CGDP) technique do?It's part of TeleOCR's geometry-aware document modeling, helping the model represent distorted or camera-captured page shapes so it can parse layout and content directly without a separate dewarping step.
What are similar projects?ZDTaichu5.0-9B, OraRL, olmocr, DeepSeek-OCR, lift, Intern-Decision, StartLux-Decision, X-AnyLabeling
Which topics is it in?vision-language-model, ocr, document-processing, reinforcement-learning, evaluation, multimodal, transformer, image-processing, inference, nvidia, computer-vision, cuda
Creator
caipeng328
Size
28.8 MB
Stars
343
Updated
CommitSep 28
Age
2mo
Commits · 90 Days