Ukr Htr Convtext

--- language: - en - uk license: apache-2.0 library_name: transformers tags: - ocr - icr - handwriting-recognition - ukrainian - ctc - pytorch - onnx pipeline_tag: image-to-text base_model: DAIR-Group/HTR-ConvText metrics: - cer - wer --- <div align="center"> # 🇺🇦 Ukrainian OCR / ICR (HTR-ConvText) **Handwritten & printed text recognition for Ukrainian** [](https://huggingface.co/spaces/Valerii02/ukr-htr-convtext-demo) *Upload an image → Get recognized text* [English](#quickstart) · [Українська](README.uk.md) </div> <div align="left" style="line-height: 1.6"> ## 📋 Table of Contents - [✨ Highlights](#highlights) - [🚀 Quickstart](#quickstart) - [📖 Model Description](#model-description) - [🖼️ Recognition Examples](#recognition-examples) - [🛠️ Tools & Scripts](#tools--scripts) - [📊 Evaluation](#evaluation) - [🙏 Attribution & Citation](#attribution--citation) ## ✨ Highlights | Feature | Description | |---------|-------------| | **Language** | Ukrainian (handwritten + printed) | | **Architecture** | HTR-ConvText (ResNet-18 + MobileViT), CTC decoding | | **Input** | 64×3072 px, grayscale line images | | **Training** | 1.7M samples, SAM, EMA, scan simulation | | **Formats** | PyTorch, ONNX, Hugging Face `AutoModel` | ## 🚀 Quickstart ```python from transformers import AutoModel, AutoProcessor processor = AutoProcessor.from_pretrained("Valerii02/ukr-htr-convtext", trust_remote_code=True) model = AutoModel.from_pretrained("Valerii02/ukr-htr-convtext", trust_remote_code=True) ``` ```python inputs = processor(images="sample.png", return_tensors="pt") logits = model(**inputs).logits text = processor.batch_decode(logits)[0] print(text) ``` > 💡 **Try it now:** [Open the Gradio demo](https://huggingface.co/spaces/Valerii02/ukr-htr-convtext-demo) — no code required! ## 📖 Model Description This repository packages a **Ukrainian OCR/ICR model** for handwritten and partially printed text with a Hugging Face–native API (`AutoModel` + `AutoProcessor`). ### Architecture - **Backbone:** ResNet-18 + MobileViT (MVP), hierarchical ConvText encoder (U-Net-like down/upsampling) - **Decoding:** CTC greedy - **Vocabulary:** 151 characters (Ukrainian + symbols) ### Training Data | Source | Samples | |--------|---------| | [ukrainian-handwriting-synth](ht...

Source context: 39 downloads · 0 likes · Pipeline image-to-text · Library transformers · Repo Valerii02/ukr-htr-convtext