Optical Character Recognition + Instance Segmentation for russian and english languages

Last update: Dec 19, 2022

Overview

Распознавание рукописного текста в школьных тетрадях

Соревнование, проводимое в рамках олимпиады НТО, разработанное Сбером. Платформа ODS.

Результаты Public

Задача

Вам нужно разработать алгоритм, который способен распознать рукописный текст в школьных тетрадях. В качестве входных данных вам будут предоставлены фотографии целых листов. Предсказание модели — список распознанных строк с координатами полигонов и получившимся текстом.

Как должно работать решение?

Последовательность двух моделей: сегментации и распознавания. Сначала сегментационная модель предсказывает полигоны маски каждого слова на фото. Затем эти слова вырезаются из изображения по контуру маски (получаются кропы на каждое слово) и подаются в модель распознавания. В итоге получается список распознанных слов с их координатами.

Модели

Instance Segmentation

модель X101-FPN из зоопарка моделей detectron2 + аугментации + высокое разрешение

Optical Character Recognition (OCR)

архитектура CRNN с бекбоном Resnet-34, предобученным на топ 1 модели соревнования Digital Peter

Beam Search

модель KenLM, обученная на данных сорвенования Feedback, Решу ОГЭ/ЕГЭ, а также CTCDecoder

Ресурсы & Submit

Christofari с NVIDIA Tesla V100 и образом jupyter-cuda10.1-tf2.3.0-pt1.6.0-gpu:0.0.82

Мы не гарантируем поддержку сабмита всё время, поэтому предоставляем 2 ссылки: Google Drive и Yandex

Цитирование

@misc{nto-ai-text-recognition,
  author =       {Arseniy Shahmatov and Gerasomiv Maxim},
  title =        {notebook-recognition},
  howpublished = {\url{https://github.com/Lednik7/nto-ai-text-recognition}},
  year =         {2022}
}

You might also like...

Mask R-CNN for object detection and instance segmentation on Keras and TensorFlow

Mask R-CNN for Object Detection and Segmentation This is an implementation of Mask R-CNN on Python 3, Keras, and TensorFlow. The model generates bound

22.5k Jan 4, 2023

This is an official implementation for "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows" on Object Detection and Instance Segmentation.

Swin Transformer for Object Detection This repo contains the supported code and configuration files to reproduce object detection results of Swin Tran

1.4k Dec 30, 2022

Fast, modular reference implementation of Instance Segmentation and Object Detection algorithms in PyTorch.

Faster R-CNN and Mask R-CNN in PyTorch 1.0 maskrcnn-benchmark has been deprecated. Please see detectron2, which includes implementations for all model

9k Jan 4, 2023

Object detection and instance segmentation toolkit based on PaddlePaddle.

9.3k Jan 2, 2023

DiscoBox: Weakly Supervised Instance Segmentation and Semantic Correspondence from Box Supervision

The Official PyTorch Implementation of DiscoBox: Weakly Supervised Instance Segmentation and Semantic Correspondence from Box Supervision

3 Oct 15, 2021

The PyTorch implementation of DiscoBox: Weakly Supervised Instance Segmentation and Semantic Correspondence from Box Supervision.

DiscoBox: Weakly Supervised Instance Segmentation and Semantic Correspondence from Box Supervision The PyTorch implementation of DiscoBox: Weakly Supe

1 Oct 23, 2021

Optical Character Recognition + Instance Segmentation for russian and english languages

Related tags

Overview

Распознавание рукописного текста в школьных тетрадях

Соревнование, проводимое в рамках олимпиады НТО, разработанное Сбером. Платформа ODS.

Результаты Public

Задача

Как должно работать решение?

Модели

Ресурсы & Submit

Цитирование

You might also like...

Mask R-CNN for object detection and instance segmentation on Keras and TensorFlow

This is an official implementation for "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows" on Object Detection and Instance Segmentation.

Fast, modular reference implementation of Instance Segmentation and Object Detection algorithms in PyTorch.

Object detection and instance segmentation toolkit based on PaddlePaddle.

DiscoBox: Weakly Supervised Instance Segmentation and Semantic Correspondence from Box Supervision

The PyTorch implementation of DiscoBox: Weakly Supervised Instance Segmentation and Semantic Correspondence from Box Supervision.

Numbering permanent and deciduous teeth via deep instance segmentation in panoramic X-rays

Res2Net for Instance segmentation and Object detection using MaskRCNN

Keras implementation of PersonLab for Multi-Person Pose Estimation and Instance Segmentation.

Releases(v1.0.0)

v1.0.0(Mar 6, 2022)

Owner

Gerasimov Maxim

This is a repository of our model for weakly-supervised video dense anticipation.

(ICCV 2021) ProHMR - Probabilistic Modeling for Human Mesh Recovery

Quasi-Dense Similarity Learning for Multiple Object Tracking, CVPR 2021 (Oral)

基于Paddle框架的arcface复现

TensorFlow implementation of ENet

Using Streamlit to host a multi-page tool with model specs and classification metrics, while also accepting user input values for prediction.

Creating Multi Task Models With Keras

Unsupervised Image-to-Image Translation

CLIPort: What and Where Pathways for Robotic Manipulation

A Python module for the generation and training of an entry-level feedforward neural network.

Efficient Conformer: Progressive Downsampling and Grouped Attention for Automatic Speech Recognition

This program can detect your face and add an Christams hat on the top of your head

Py4fi2nd - Jupyter Notebooks and code for Python for Finance (2nd ed., O'Reilly) by Yves Hilpisch.

The easiest tool for extracting radiomics features and training ML models on them.

Unsupervised Discovery of Object Radiance Fields

An ever-growing playground of notebooks showcasing CLIP's impressive zero-shot capabilities.

PyAF is an Open Source Python library for Automatic Time Series Forecasting built on top of popular pydata modules.

Official implementation of EdiTTS: Score-based Editing for Controllable Text-to-Speech

Code for "Contextual Non-Local Alignment over Full-Scale Representation for Text-Based Person Search"

Predicts an answer in yes or no.