search

ariaaay / clip2brain

Code from paper "Natural language supervision with a large and diverse dataset builds better models of human high-level visual cortex"
MIT License
14 stars 6 forks source link

About the images #3

Closed Brickkkkkk closed 7 months ago

Brickkkkkk commented 9 months ago

Hello! I have some problems when I run the codes. What's the format of NSD_images and where can I download them?

ariaaay commented 8 months ago

They should be available when you download the dataset from https://naturalscenesdataset.org/.

geoffreyfan commented 8 months ago

image Hello! The names of the 10,000 images being read from within are based on specific IDs. How can I obtain these 10,000 images that conform to the required code ID format?

JustMuteAll commented 8 months ago

These specific IDs are their IDs in COCO image dataset,instead of IDs in NSD dataset. An alternative solution is downloading the whole COCO 2017 train/val image dataset and set the COCO dataset as the stimuli folder.