训练数据共约20w通用分割粗标注数据。 该模型在fss1000数据测试集上zero shot iou: 85.14.文本指导的分割模型
模型结构信息
使用方式和范围
如何使用
代码范例
# Copyright (c) Alibaba, Ic. ad its affiliates.
from modelscope.outputs import OutputKeys
from modelscope.pipelies import pipelie
from modelscope.utils.costat import Tasks
iput_locatio = 'https://clip-multimodal.oss-c-beijig.aliyucs.com/xigguag/maas/data/text_drive_seg_demo.jpg'
test_iput = {
'image': iput_locatio,
'text': 'bear',
}
model_id = 'damo/cv_vitl16_segmetatio_text-drive-seg'
shop_seg = pipelie(Tasks.text_drive_segmetatio, model=model_id)
result = shop_seg(test_iput)
import cv2
# result[OutputKeys.MASKS] is segmet map result,other keys are ot used
cv2.imwrite('demo_textdriveseg.jpg', result[OutputKeys.MASKS])
使用场景
模型局限性以及可能的偏差
训练数据
模型训练
预处理
数据评估及结果
相关论文以及引用信息
@iproceedigs{
li2022laguagedrive,
title={Laguage-drive Sematic Segmetatio},
author={Boyi Li ad Kilia Q Weiberger ad Serge Belogie ad Vladle Koltu ad Ree Raftl},
booktitle={Iteratioal Coferece o Learig Represetatios},
year={2022},
url={https://opereview.et/forum?id=RriDjddCLN}
}
点击空白处退出提示







评论