MSPE: Multi-Scale Patch Embedding Prompts Vision Transformers to Any Resolution.
Wenzhuo LiuFei ZhuShijie MaCheng-Lin LiuPublished in: CoRR (2024)
Keyphrases
- multiscale
- computer vision
- high resolution
- low resolution
- image processing
- image patches
- natural images
- vision system
- edge detection
- coarse to fine
- consequence finding
- real time
- nonlinear dimensionality reduction
- multiple scales
- visual perception
- vector space
- scale space
- keypoints
- motion estimation
- feature vectors
- control group
- mental models
- wavelet decomposition
- watermarking algorithm
- learning environment
- high quality
- multidimensional scaling
- higher resolution
- image segmentation
- graph embedding
- data sets