Skip to content
Other

anaksor-ocr

OCR для кхмерского и латиницы

2 min read

Sokheng выложил веса OCR для кхмерского и английского/латинского текста на документах, чеках и фотографиях вывесок. Это пара моделей: детектор Mobile DBNet++ и распознаватель SVTRv2-B + CTC; текущие версии — DBNet++ v1 и rec v4. Есть Python-пакет anaksor-ocr с ONNX Runtime или PyTorch.

Ключевые особенности

  • Детектор Mobile DBNet++: RepViT backbone + RSEFPN + Adaptive Scale Fusion + DB head — находит строки текста.
  • Распознаватель SVTRv2-B с CTC-головой читает каждую вырезанную строку; в конфиге высота 48 и до 140 символов.
  • Пакет работает через ONNX Runtime без PyTorch или с PyTorch-бэкендом (CUDA, Apple MPS).
  • Можно закрепить релиз (rec-v4-det-dbpp-v1) или запускать офлайн с локальными весами.
  • Загрузка проверяет SHA-256; форматы .onnx/.safetensors не исполняют код при загрузке.
  • Набор символов: 211 символов + пробел; поддержка кхмерского и латиницы.

Пример

import anaksor

ocr = anaksor.OCR()
result = ocr("letter.jpg")
print(result["text"])
ПодробнееБенчмарки: rec v4 на real validation — 86.7% line accuracy / 1.38% CER; Latin — 76.4% / 4.29%. Detection на HierText — hmean 0.724; на повороте 45° — 0.421 против 0.303 у v3.
Детектор: 3.15M параметров (+21k к v3), дообучен 40 эпох.
Обучение rec: fine-tune от китайского SVTRv2-B из OpenOCR на KhmerST, HierText и 300k синтетических строк.
Ограничения: в основном печатный кхмерский; рукописный текст и декоративные шрифты слабее; латиница на фото — самый слабый случай. Длинные строки читаются перекрывающимися окнами.
Лицензия весов — Apache-2.0.
Discuss Telegram 1

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.1 GB · ready files0.6 GB
  • FP8 / INT8recommendedweights 0.0 GB0.5 GB
  • 4-bitweights 0.0 GB0.5 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · BF16 — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # Sokheng/anaksor-ocr
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download Sokheng/anaksor-ocr --local-dir ~/models/anaksor-ocr
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to anaksor-ocr
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of anaksor-ocr (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

Sokheng/anaksor-ocr
Author
Sokheng
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

01

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments

Other

OpenHeart

OpenHeart