metadata

license: other
dataset_info:
  features:
    - name: subject
      dtype: string
    - name: subject_id
      dtype: int64
    - name: nsd_id
      dtype: int64
    - name: image
      dtype: image
    - name: activity
      dtype: image
    - name: coco_split
      dtype: string
    - name: coco_id
      dtype: int64
    - name: objects
      struct:
        - name: area
          sequence: int64
        - name: bbox
          sequence:
            sequence: float64
        - name: category
          sequence: string
        - name: iscrowd
          sequence: int64
        - name: segmentation
          list:
            - name: counts
              dtype: string
            - name: poly
              sequence:
                sequence: float64
            - name: size
              sequence: int64
        - name: supercategory
          sequence: string
        - name: target
          sequence: int64
    - name: captions
      sequence: string
  splits:
    - name: train
      num_bytes: 26654549442
      num_examples: 195000
  download_size: 12966047661
  dataset_size: 26654549442
task_categories:
  - image-to-image
  - object-detection
tags:
  - biology
  - neuroscience
  - fmri
size_categories:
  - 100K<n<1M

NSD-Flat

[GitHub] [🤗 Hugging Face Hub]

A Hugging Face dataset of pre-processed brain activity flat maps from the Natural Scenes Dataset, constrained to a visual cortex region of interest and rendered as PNG images.

Load the dataset

Load the dataset from Hugging Face Hub

from datasets import load_dataset

dataset = load_dataset("clane9/NSD-Flat", split="train")

Building the dataset

1. Download source data

Run download_data.sh to download the required source data:

NSD stimuli images and presentation info
COCO annotations
NSD beta activity maps in fsaverge surface space

bash download_data.sh

2. Convert the COCO annotations

Run convert_nsd_annotations.py to crop and reorganize the COCO annotations for NSD.

python convert_nsd_annotations.py

3. Generate the dataset

Run generate_dataset.py to generate the huggingface dataset in Arrow format.

python generate_dataset.py --img_size 256 --workers 8

Citation

If you find this dataset useful, please consider citing:

@article{allen2022massive,
  title     = {A massive 7T fMRI dataset to bridge cognitive neuroscience and artificial intelligence},
  author    = {Allen, Emily J and St-Yves, Ghislain and Wu, Yihan and Breedlove, Jesse L and Prince, Jacob S and Dowdle, Logan T and Nau, Matthias and Caron, Brad and Pestilli, Franco and Charest, Ian and others},
  journal   = {Nature neuroscience},
  volume    = {25},
  number    = {1},
  pages     = {116--126},
  year      = {2022},
  publisher = {Nature Publishing Group US New York}
}

@misc{lane2023nsdflat,
  author       = {Connor Lane},
  title        = {NSD-Flat: Pre-processed brain activity flat maps from the Natural Scenes Dataset},
  howpublished = {\url{https://huggingface.co/datasets/clane9/NSD-Flat}},
  year         = {2023},
}

License

Usage of this dataset constitutes agreement to the NSD Terms and Conditions.