search

Aldrich2y / MIANet

Official PyTorch Implementation of MIANet: Aggregating Unbiased Instance and General Information for Few-Shot Semantic Segmentation(CVPR 2023).
27 stars 2 forks source link

Word embedding #8

Open hy0523 opened 1 year ago

hy0523 commented 1 year ago

Hello, thanks for you wonderful work! I would like to ask how to generate the word embeddings in your paper, can you provide some instructions or code for that?

Aldrich2y commented 6 months ago

As you can see in the paper and in the file embeddings/coco.txt.

  1. A single category is fed directly into the word vector model to get a 300-dimensional result.
    1. If a category contains multiple words, they will be sent to the word vector model separately, and the average value will be used as the result, such as traffic and light; stop and sign.
    2. All output results in each dataset will eventually be normalized.
    3. The complete instructions or code is being compiled soon.