search

yongliu20 / UniLSeg

[CVPR 2024] Official implementation of "Universal Segmentation at Arbitrary Granularity with Language Instruction"
MIT License
71 stars 0 forks source link

分割图像中的所有物体 #10

Closed BaiCHao12 closed 1 month ago

BaiCHao12 commented 1 month ago

注意到在bpe_simple_vocab_16e6.txt.gz文件中没有发现object、all objects等提示词，想分割图像中的所有物体 提示词写什么

yongliu20 commented 1 month ago

UniLSeg is designed to segment the target region based on the specific semantics of input text. The examples of segmenting all objects in the paper are achieved by combining different text inputs. Therefore, if you want to segment all the objects present in the image, you need to input different text prompts and combine the segmentation results.