deepfates / memery

Search over large image datasets with natural language and computer vision!
https://deepfates.com/memery
MIT License
525 stars 27 forks source link

image as input #5

Closed deepfates closed 3 years ago

deepfates commented 3 years ago

CLIP encodes images and text to the same latent space. It should be easy to build an input for images, and either search purely on the image or on an image+text vector combination

deepfates commented 3 years ago

Added in 678cd5e7c2864fa6fb7c3dff5516227d301f73a4