from Hacker News

EfficientSAM

by Thomashuet on 12/6/23, 12:35 PM with 7 comments

• by skadamat on 12/6/23, 6:25 PM

  Excited to play with this more! Forked the repo and added the models into the repo itself (migrated from Dropbox): https://github.com/xetdata/EfficientSAM

• by IshanMi on 12/6/23, 9:34 PM

  So if I'm understanding this correctly:

  The SAM paper from this past April (that let you do zero-shot segmentation on any image, seemingly better than even OpenAI's CLIP) was using a ~600M parameter ViT model to generate image embeddings. And in order to make it less computationally expensive to generate those same embeddings, they replace that model with a smaller ViT encoder that was pre-trained using the masked auto-encoder back propagation method?

• by GaggiX on 12/6/23, 1:38 PM

  https://github.com/ChaoningZhang/MobileSAM was the previous attempt at reducing the size of the large image encoder used by SAM.

• by cchance on 12/6/23, 5:13 PM

  it's called efficient Sam and it appears to be onpar or better than fastsam but did I miss a memory or speed comparison?

• by naveen99 on 12/6/23, 1:25 PM

  can’t wait for everywhere all at once function.

• by ShadowBanThis01 on 12/6/23, 8:46 PM

  Is what?