AICUP (隊伍: CCUML_頂樓集合)

Introduction

場景文字檢測通常為場景文字辨識的前置步驟，即由畫面的像素中判斷文字出現的位置，以利後續針對該位置辨識可能的文字內容。場景文字檢測直接影響了文字辨識的準確度，本次賽事目標即為定位畫面中肉眼可識的文字位置。場景文字檢測受到許多因素所影響，包括場景中可能出現的多型態文字、多國文字、傾斜招牌文字、不同尺寸文字、外物遮蔽、類文字圖案紋理干擾、光線與陰影等。本賽事的目標場景為台灣市區街景，期望參賽者利用機器學習/深度學習技術，嘗試與開發適當的模型，以確偵測台灣街景畫面中的文字區域。

Configuration Environment

Linux Ubuntu 18.04.04 LTS
Python 3.7
CUDA version: 10.1, V10.1.243

Folder Structure

dataset dir:

Training and validation data are placed here.

|__AICUP2021
|  |__TrainDataset
|  |  |__img
|  |  |  |__img_1.jpg
|  |  |__train_gts
|  |  |  |__img_1.txt
|  |  |__train_list.txt
|  |__ValidationDataset
|  |  |__img
|  |  |  |__img_3701.jpg
|  |  |__valid_gts
|  |  |  |__img_3701.txt
|  |  |__valid_list.txt

experiments dir:

The YAML file with the name of *AICUP*.yaml, used to define the model structure 、 training hyperparameters and data preprocessing.

|__seg_detector
|  |__base.yaml
|  |__base_AICUP.yaml
|  |__AICUP_resnet50_deform_thre.yaml

Installation requirements library

pip install -r requirements.txt

Running experiments

Training

CUDA_VISIBLE_DEVICES=0 python train.py {config_yaml} \
--num_gpus 1 --epochs 1 --num_workers 4 --batch_size 8

Evaluate

CUDA_VISIBLE_DEVICES=0 python eval.py {config_yaml} --resume {model_weight_path} --box_thresh 0.5

Predict

CUDA_VISIBLE_DEVICES=0 python origiDemo.py {config_yaml} \
--resume {model_weight_path}} \
--image_path {image_dir_path} \
--box_thresh 0.5 --visualize

Generate Excel Result

CUDA_VISIBLE_DEVICES=0 python demo.py {config_yaml} \
--resume {model_weight_path}} \
--image_path {image_dir_path} \
--box_thresh 0.5 \
--visualize \
--output_csvName {csv_name}

Reference

Liao, M.; Wan, Z.; Yao, C.; Chen, K.; and Bai, X. 2020. Real-Time Scene Text Detection with Differentiable Binarization. In AAAI.
He, K., Zhang, X., Ren, S., Sun, J.: Deep Residual Learning for Image Recognition. CoRR abs/1512.03385 (2015)
G. Huang, Z. Liu, K. Q. Weinberger, and L. Maaten. Densely connected convolutional networks. In CVPR, 2017

Acknowledgment

Most of the program code are based on DB(Github).

Members

Chen-Mao Liao

Hong-Jia Chen

Yu-Ruei Lin

wilile26811249/AICUP_DB