给定一张输入图像,输出视觉显著注意力图。
本模型适用范围较广,预测像素视觉显著注意程度,但不涉及图像中的语义信息。
在ModelScope框架上,提供输入图片,即可以通过简单的Pipeline调用使用当前模型。
from modelscope.pipelines import pipeline
from modelscope.utils.constant import Tasks
from modelscope.outputs import OutputKeys
salient_detect = pipeline(Tasks.semantic_segmentation, model='damo/cv_u2net_salient-detection')
img_path ='https://modelscope.oss-cn-beijing.aliyuncs.com/test/images/image_salient_detection.jpg'
result = salient_detect(img_path)
import cv2
cv2.imwrite('./result.jpg',result[OutputKeys.MASKS])
DataSet | MAE | Sm | maxF~β | F^w~β |
---|---|---|---|---|
DUTS-TE | 0.044 | 0.861 | 0.873 | 0.804 |
DUT-OMRON | 0.054 | 0.847 | 0.823 | 0.757 |
SOD | 0.108 | 0.786 | 0.861 | 0.748 |
ECSSD | 0.033 | 0.928 | 0.951 | 0.910 |
HKU-IS | 0.031 | 0.916 | 0.935 | 0.890 |
@article{Qin_2020_PR,
title = {U2-Net: Going Deeper with Nested U-Structure for Salient Object Detection},
author = {Qin, Xuebin and Zhang, Zichen and Huang, Chenyang and Dehghan, Masood and Zaiane, Osmar and Jagersand, Martin},
journal = {Pattern Recognition},
volume = {106},
pages = {107404},
year = {2020}
}