Arbitrary Reading Order Scene Text Spotter with Local Semantics Guidance

News | Installation | Datasets | Training | Inference | Acknowledgement

This is the official repo for the paper "Arbitrary Reading Order Scene Text Spotter with Local Semantics Guidance", which is accepted to AAAI 2025.

We will release our code soon.

News

2024/12/19

• Update the README!

2024/12/15

• Update the arxiv version!

2024/12/10

• The paper is accepted by AAAI 2025!

Installation

conda create -n lsg python=3.9 -y
conda activate lsg
pip install torch==1.10.0+cu111 torchvision==0.11.0+cu111 torchaudio==0.10.0 -f https://download.pytorch.org/whl/torch_stable.html
pip install -U openmim
mim install mmcv-full==1.5.2
mim install mmdet==2.28.2
git clone https://github.com/pd162/LSG
cd LSG
pip install -e .

Datasets

The convert annotations can be download from [Google Drive], Please download and extract the above datasets into the data folder following the file structure below.

data
├─totaltext
│  │ totaltext_train.json
│  │ totaltext_test.json
│  └─imgs
│      ├─training
│      └─test
├─CTW1500
│  │ instances_training.json
│  │ instance_test.json
│  └─imgs
│      ├─training
│      └─test
├─mlt
│  │  train_polygon.json
│  └─images
├─synthtext-150k
      ├─syntext1
      │  │  train_polygon.json
      │  └─images
      ├─syntext2
         │  train_polygon.json
         └─images

Training

CUDA_VISIBLE_DEVICES=0,1 ./tools/train.sh config/LSG/lsg_pretrain.py work_dirs/pretrain 2

Inference

CUDA_VISIBLE_DEVICES=0 python tools/test.py config/tpsnet/tpsnet_totaltext.py work_dirs/totaltext/latest.pth --eval eval_text

Acknowledgement

We sincerely thank MMOCR, TPSNet for their excellent works.