Masked Image Pretraining on Language Assisted Representation

Keyphrases

• Mask Image 100%
• Masked Autoencoder 100%
• Image Features 50%
• 4-point 50%
• Labeled Data 50%
• Sampling Mechanism 50%
• Semantic Segmentation 50%
• Quality Model 50%
• Encoder-decoder Architecture 50%
• Low-level Features 50%
• Pixel Features 50%
• Computer Vision Tasks 50%
• Reconstruction-based 50%
• Input Resolution 50%
• Transformer Model 50%
• Semantic-aware 50%
• Dataset Reconstruction 50%
• Mask Sampling 50%
• Image Dataset 50%
• Weak Supervision 50%
• Unlabeled Images 50%
• Self-attention 50%
• Transfer Performance 50%
• Labeled Images 50%
• Cross-attention Transformer 50%
• ImageNet-1K 50%
• Transferable Representation 50%
• ADE20K 50%
• Masked Pretraining 50%
• Text Caption 50%
• MIoU 50%
• Language Supervision 50%

Computer Science

• Autoencoder 100%
• Experimental Result 50%
• image feature 50%
• Image Segmentation 50%
• Segmentation Task 50%
• Computer Vision Task 50%
• quality model 50%
• Transformer Model 50%
• Self-Attention 50%