Efficient and Accurate Arbitrary-Shaped Text Detection with Pixel Aggregation Network

Overview

Efficient and Accurate Arbitrary-Shaped Text Detection with Pixel Aggregation Network

Paddle-PANet

目录

结果对比

CTW1500

Method Backbone Fine-tuning Config Precision (%) Recall (%) F-measure (%) Model Log
mmocr_PANet Resnet18 N ctw_config 77.6 83.8 80.6 -- --
PAN (paper) ResNet18 N config 84.6 77.7 81.0 - -
PaddlePaddle_PANet ResNet18 N panet_r18_ctw.py 84.51 78.62 81.46 Model Log

论文介绍

背景简介

这是发在2019ICCV上的一篇一阶段场景文本检测论文。主要是PSENet的升级版。PSENet虽然处理速度很快,准确度很高,但后处理过程繁琐,而且没办法和网络模型融合在一起,实现训练。PANet很好的解决了这一问题,把后处理过程也放入网络中,预测出三个loss,最后进行融合。

网络结构

上图为PAN的整个网络结构,网络主要由Backbone + Segmentation Head(FPEM + FFM) + Output(Text Region、Kernel、Similarity Vector)组成。

本文使用ResNet-18作为PAN的默认Backbone,并提出了低计算量的Segmentation Head(FPFE + FFM)以解决因为使用ResNet-18而导致的特征提取能力较弱,特征感受野较小且表征能力不足的缺点。

此外,为了精准地重建完整的文字实例(text instance),提出了一个可学习的后处理方法——像素聚合法(PA),它能够通过预测出的相似向量来引导文字像素聚合到正确的kernel上去。

下面将详细介绍一下上面的各个部分。

Backbone

Backbone选择的是resnet18, 提取stride为4,8,16,32的conv2,conv3,conv4,conv5的输出作为高低层特征。每层的特征图的通道数都使用1*1卷积降维至128得到轻量级的特征图Fr。

Segmentation Head

PAN使用resNet-18作为网络的默认backbone,虽减少了计算量,但是backbone层数的减少势必会带来模型学习能力的下降。为了提高效率,作者在 resNet-18基础上提出了一个低计算量但可高效增强特征的分割头Segmentation Head。它由两个关键模块组成:特征金字塔增强模块(Feature Pyramid Enhancement Module,FPEM)、特征融合模块(Feature Fusion Module,FFM)。

FPEM

Feature Pyramid Enhancement Module(FPEM),即特征金字塔增强模块。FPEM呈级联结构且计算量小,可以连接在backbone后面让不同尺寸的特征更深、更具表征能力,结构如下:

FPEM是一个U形模组,由两个阶段组成,up-scale增强、down-scale增强。up-scale增强作用于输入的特征金字塔,它以步长32,16,8,4像素在特征图上迭代增强。在down-scale阶段,输入的是由up-scale增强生成的特征金字塔,增强的步长从4到32,同时,down-scale增强输出的的特征金字塔就是最终FPEM的输出。 FPEM模块可以看成是一个轻量级的FPN,只不过这个FPEM计算量不大,可以不停级联以达到不停增强特征的作用。

FFM

Feature Fusion Module(FFM)模块用于融合不同尺度的特征,其结构如下:

最后通过上采样将它们Concatenate到一起。

模型最后预测三种信息: 1、文字区域 2、文字kernel 3、文字kernel的相似向量

Loss

其中文字区域和kernel预测loss为:

快速安装

Recommended environment

Python 3.6+
paddlepaddle-gpu 2.0.2
nccl 2.0+
mmcv 0.2.12
editdistance
Polygon3
pyclipper
opencv-python 3.4.2.17
Cython

Install env

Install paddle following the official tutorial.

pip install -r requirement.txt
./compile.sh

Dataset

Please refer to dataset/README.md for dataset preparation.

Pretrain Backbone

download resent18 pre-train model in pretrain/resnet18.pdparams

pretrain_resnet18 password: j5g3

Training

CUDA_VISIBLE_DEVICES=0,1,2,3 python dist_train.py ${CONFIG_FILE}

For example:

CUDA_VISIBLE_DEVICES=0,1,2,3 python dist_train.py config/pan/pan_r18_ctw.py
#checkpoint continue
python3.7 dist_train.py config/pan/pan_r18_ctw_train.py --nprocs 1 --resume checkpoints/pan_r18_ctw_train

Evaluation

The evaluation scripts of CTW 1500 dataset. CTW

Text detection

./start_test.sh

License

This project is developed and maintained by IMAGINE [email protected] Key Laboratory for Novel Software Technology, Nanjing University.

This project is released under the Apache 2.0 license.

@inproceedings{wang2019efficient,
  title={Efficient and Accurate Arbitrary-Shaped Text Detection with Pixel Aggregation Network},
  author={Wang, Wenhai and Xie, Enze and Song, Xiaoge and Zang, Yuhang and Wang, Wenjia and Lu, Tong and Yu, Gang and Shen, Chunhua},
  booktitle={Proceedings of the IEEE International Conference on Computer Vision},
  pages={8440--8449},
  year={2019}
}
Owner
Dreams Are Messages From The Deep🪐
On Out-of-distribution Detection with Energy-based Models

On Out-of-distribution Detection with Energy-based Models This repository contains the code for the experiments conducted in the paper On Out-of-distr

Sven 19 Aug 07, 2022
KSAI Lite is a deep learning inference framework of kingsoft, based on tensorflow lite

KSAI Lite is a deep learning inference framework of kingsoft, based on tensorflow lite

80 Dec 27, 2022
GPU Accelerated Non-rigid ICP for surface registration

GPU Accelerated Non-rigid ICP for surface registration Introduction Preivous Non-rigid ICP algorithm is usually implemented on CPU, and needs to solve

Haozhe Wu 144 Jan 04, 2023
An example showing how to use jax to train resnet50 on multi-node multi-GPU

jax-multi-gpu-resnet50-example This repo shows how to use jax for multi-node multi-GPU training. The example is adapted from the resnet50 example in d

Yangzihao Wang 20 Jul 04, 2022
Predictive Maintenance LSTM

Predictive-Maintenance-LSTM - Predictive maintenance study for Complex case study, we've obtained failure causes by operational error and more deeply by design mistakes.

Amir M. Sadafi 1 Dec 31, 2021
Code for the ICCV'21 paper "Context-aware Scene Graph Generation with Seq2Seq Transformers"

ICCV'21 Context-aware Scene Graph Generation with Seq2Seq Transformers Authors: Yichao Lu*, Himanshu Rai*, Cheng Chang*, Boris Knyazev†, Guangwei Yu,

Layer6 Labs 37 Dec 18, 2022
This is the official code for the paper "Learning with Nested Scene Modeling and Cooperative Architecture Search for Low-Light Vision"

RUAS This is the official code for the paper "Learning with Nested Scene Modeling and Cooperative Architecture Search for Low-Light Vision" A prelimin

Vision & Optimization Group (VOG) 2 May 05, 2022
Convolutional Neural Networks

Darknet Darknet is an open source neural network framework written in C and CUDA. It is fast, easy to install, and supports CPU and GPU computation. D

Joseph Redmon 23.7k Jan 05, 2023
Deep Learning pipeline for motor-imagery classification.

BCI-ToolBox 1. Introduction BCI-ToolBox is deep learning pipeline for motor-imagery classification. This repo contains five models: ShallowConvNet, De

DongHee 18 Oct 31, 2022
JAXMAPP: JAX-based Library for Multi-Agent Path Planning in Continuous Spaces

JAXMAPP: JAX-based Library for Multi-Agent Path Planning in Continuous Spaces JAXMAPP is a JAX-based library for multi-agent path planning (MAPP) in c

OMRON SINIC X 24 Dec 28, 2022
tensorrt int8 量化yolov5 4.0 onnx模型

onnx模型转换为 int8 tensorrt引擎

123 Dec 28, 2022
Implementation of 🦩 Flamingo, state-of-the-art few-shot visual question answering attention net out of Deepmind, in Pytorch

🦩 Flamingo - Pytorch Implementation of Flamingo, state-of-the-art few-shot visual question answering attention net, in Pytorch. It will include the p

Phil Wang 630 Dec 28, 2022
Pytorch implementation of ICASSP 2022 paper Attention Probe: Vision Transformer Distillation in the Wild

Attention Probe: Vision Transformer Distillation in the Wild Jiahao Wang, Mingdeng Cao, Shuwei Shi, Baoyuan Wu, Yujiu Yang In ICASSP 2022 This code is

IIGROUP 6 Sep 21, 2022
Code and Resources for the Transformer Encoder Reasoning Network (TERN)

Transformer Encoder Reasoning Network Code for the cross-modal visual-linguistic retrieval method from "Transformer Reasoning Network for Image-Text M

Nicola Messina 53 Dec 30, 2022
A series of Jupyter notebooks with Chinese comment that walk you through the fundamentals of Machine Learning and Deep Learning in python using Scikit-Learn and TensorFlow.

Hands-on-Machine-Learning 目的 这份笔记旨在帮助中文学习者以一种较快较系统的方式入门机器学习, 是在学习Hands-on Machine Learning with Scikit-Learn and TensorFlow这本书的 时候做的个人笔记: 此项目的可取之处 原书的

Baymax 1.5k Dec 21, 2022
Understanding Hyperdimensional Computing for Parallel Single-Pass Learning

Understanding Hyperdimensional Computing for Parallel Single-Pass Learning Authors: Tao Yu* Yichi Zhang* Zhiru Zhang Christopher De Sa *: Equal Contri

Cornell RelaxML 4 Sep 08, 2022
Implementation of "GNNAutoScale: Scalable and Expressive Graph Neural Networks via Historical Embeddings" in PyTorch

PyGAS: Auto-Scaling GNNs in PyG PyGAS is the practical realization of our G NN A uto S cale (GAS) framework, which scales arbitrary message-passing GN

Matthias Fey 139 Dec 25, 2022
Live Hand Tracking Using Python

Live-Hand-Tracking-Using-Python Project Description: In this project, we will be

Hassan Shahzad 2 Jan 06, 2022
DANet for Tabular data classification/ regression.

Deep Abstract Networks A pyTorch implementation for AAAI-2022 paper DANets: Deep Abstract Networks for Tabular Data Classification and Regression. Bri

Ronnie Rocket 55 Sep 14, 2022
Calibrate your listeners! Robust communication-based training for pragmatic speakers. Findings of EMNLP 2021.

Calibrate your listeners! Robust communication-based training for pragmatic speakers Rose E. Wang, Julia White, Jesse Mu, Noah D. Goodman Findings of

Rose E. Wang 3 Apr 02, 2022