Datasets:

mcosarinsky
/

NSD-VQA

image_id int64 0 123k	question_id int64 1 2.1M	answer_id int64 1 2.1M	question stringlengths 20 48	answers listlengths 1 7	category stringclasses 25 values
0	1	1	Is this scene indoor or outdoor?	[ "outdoor" ]	scene
0	2	2	Where is this scene located?	[ "beach" ]	location
0	3	3	What is the person doing?	[ "surfing" ]	action
0	4	4	What is the person holding?	[ "surfboard" ]	holding
0	5	5	Is person in the image?	[ "yes" ]	person yes/no
0	6	6	Is surfboard in the image?	[ "yes" ]	sport yes/no
0	7	7	Is water in the image?	[ "yes" ]	nature yes/no
0	9	9	What color is the surfboard?	[ "white" ]	color
0	10	10	What color is the water?	[ "blue" ]	color
0	11	11	How many people are in this image?	[ "1" ]	counting
0	12	12	How many surfboards are in this image?	[ "1" ]	counting
1	13	13	Is this scene indoor or outdoor?	[ "outdoor" ]	scene
1	14	14	Where is this scene located?	[ "water" ]	location
1	15	15	Which animal is in the image?	[ "polar bear" ]	animal
1	16	16	Is polar bear in the image?	[ "yes" ]	animal yes/no
1	18	18	How many polar bears are in this image?	[ "1" ]	counting
2	19	19	Is this scene indoor or outdoor?	[ "indoor" ]	scene
2	20	20	Where is this scene located?	[ "room" ]	location
2	21	21	Is doll in the image?	[ "yes" ]	objects yes/no
2	23	23	Is basket in the image?	[ "yes" ]	household yes/no
2	24	24	What color is the doll?	[ "white" ]	color
2	26	26	What color is the basket?	[ "brown" ]	color
2	27	27	How many dolls are in this image?	[ "1" ]	counting
2	28	28	How many teddy bears are in this image?	[ "1" ]	counting
2	29	29	How many baskets are in this image?	[ "1" ]	counting
3	30	30	Is this scene indoor or outdoor?	[ "outdoor" ]	scene
3	31	31	Where is this scene located?	[ "street" ]	location
3	32	32	Is building in the image?	[ "yes" ]	structure yes/no
3	33	33	Is clock in the image?	[ "yes" ]	objects yes/no
3	34	34	Is statue in the image?	[ "yes" ]	structure yes/no
3	35	35	Is flag in the image?	[ "yes" ]	other yes/no
3	36	36	What color is the building?	[ "red" ]	color
3	37	37	What color is the clock?	[ "white" ]	color
3	38	38	What color is the statue?	[ "brown" ]	color
3	39	39	What color is the flag?	[ "red" ]	color
3	40	40	How many buildings are in this image?	[ "1" ]	counting
3	41	41	How many clocks are in this image?	[ "1" ]	counting
3	42	42	How many statues are in this image?	[ "3" ]	counting
3	43	43	How many flags are in this image?	[ "1" ]	counting
4	44	44	Is this scene indoor or outdoor?	[ "outdoor" ]	scene
4	45	45	Where is this scene located?	[ "garden" ]	location
4	46	46	Is table in the image?	[ "yes" ]	household yes/no
4	47	47	Is bench in the image?	[ "yes" ]	household yes/no
4	48	48	What color is the table?	[ "brown" ]	color
4	49	49	What color is the bench?	[ "brown" ]	color
4	50	50	How many tables are in this image?	[ "1" ]	counting
4	51	51	How many benches are in this image?	[ "2" ]	counting
5	52	52	Is this scene indoor or outdoor?	[ "outdoor" ]	scene
5	53	53	Where is this scene located?	[ "street" ]	location
5	54	54	Which vehicle is in the image?	[ "car" ]	vehicle
5	55	55	Is flag in the image?	[ "yes" ]	other yes/no
5	57	57	Is tree in the image?	[ "yes" ]	nature yes/no
5	58	58	Is street light in the image?	[ "yes" ]	street yes/no
5	59	59	Is car in the image?	[ "yes" ]	vehicle yes/no
5	60	60	What color is the flag?	[ "red" ]	color
5	61	61	What color is the building?	[ "blue" ]	color
5	62	62	What color is the tree?	[ "green" ]	color
5	63	63	What color is the street light?	[ "black" ]	color
5	64	64	What color is the car?	[ "red" ]	color
5	65	65	How many flags are in this image?	[ "few" ]	counting
5	66	66	How many buildings are in this image?	[ "2" ]	counting
5	67	67	How many trees are in this image?	[ "many" ]	counting
5	68	68	How many street lights are in this image?	[ "1" ]	counting
5	69	69	How many cars are in this image?	[ "1" ]	counting
6	70	70	Is this scene indoor or outdoor?	[ "indoor" ]	scene
6	71	71	Where is this scene located?	[ "bathroom" ]	location
6	72	72	Is toilet in the image?	[ "yes" ]	household yes/no
6	73	73	Is toilet paper in the image?	[ "yes" ]	household yes/no
6	74	74	Is tissue box in the image?	[ "yes" ]	household yes/no
6	77	77	What color is the tissue box?	[ "brown" ]	color
6	78	78	How many toilets are in this image?	[ "1" ]	counting
6	79	79	How many toilet papers are in this image?	[ "1" ]	counting
6	80	80	How many tissue boxes are in this image?	[ "1" ]	counting
7	81	81	Is this scene indoor or outdoor?	[ "outdoor" ]	scene
7	82	82	Where is this scene located?	[ "park" ]	location
7	83	83	What is the person doing?	[ "skateboarding" ]	action
7	84	84	Is person in the image?	[ "yes" ]	person yes/no
7	85	85	Is skateboard in the image?	[ "yes" ]	sport yes/no
7	86	86	Is bench in the image?	[ "yes" ]	household yes/no
7	87	87	Is tree in the image?	[ "yes" ]	nature yes/no
7	90	90	What color is the bench?	[ "gray" ]	color
7	91	91	What color is the tree?	[ "green" ]	color
7	92	92	How many people are in this image?	[ "1" ]	counting
7	93	93	How many skateboards are in this image?	[ "1" ]	counting
7	94	94	How many benches are in this image?	[ "1" ]	counting
7	95	95	How many trees are in this image?	[ "many" ]	counting
8	96	96	Is this scene indoor or outdoor?	[ "indoor" ]	scene
8	97	97	Where is this scene located?	[ "event" ]	location
8	98	98	What is the person doing?	[ "sitting" ]	action
8	99	99	Is person in the image?	[ "yes" ]	person yes/no
8	100	100	Is chair in the image?	[ "yes" ]	household yes/no
8	102	102	What color is the chair?	[ "white" ]	color
8	103	103	How many people are in this image?	[ "2" ]	counting
8	104	104	How many chairs are in this image?	[ "2" ]	counting
9	105	105	Is this scene indoor or outdoor?	[ "outdoor" ]	scene
9	106	106	Where is this scene located?	[ "lake" ]	location
9	107	107	What is the person doing?	[ "sitting" ]	action
9	108	108	Which vehicle is in the image?	[ "boat" ]	vehicle
9	109	109	Is person in the image?	[ "yes" ]	person yes/no
9	110	110	Is boat in the image?	[ "yes" ]	vehicle yes/no

End of preview. Expand in Data Studio

NSD-VQA

NSD-VQA is a large-scale visual question answering benchmark for studying what visual and semantic information can be decoded from human fMRI responses to natural images.

It is introduced in the paper Brain-IT-VQA: From Brain Signals to Answers.

🔗 Project page: https://matiascosarinsky.github.io/brain-it-vqa/

Overview

NSD-VQA is built from the Natural Scenes Dataset (NSD) and provides automatically generated question-answer annotations grounded in NSD images.

The NSD candidate-policy variant contains approximately:

73K NSD images
~123K external COCO images
~20 question-answer pairs per image
20 controlled semantic question categories

Question categories include:

object presence
counting
color
actions
scene understanding
human-object interactions
semantic categories such as animals, vehicles, food, location, etc.

This repository contains only generated annotations and metadata. It does not redistribute NSD images, external source images, or fMRI recordings. Users must obtain the original source datasets separately under their original terms.

Dataset Variants

File	Description
`nsd_vqa.parquet`	Candidate-policy NSD-VQA annotations with short answers (1,080,457 pairs)
`nsd_vqa_fs.parquet`	Candidate-policy full-sentence answer variant, NSD-VQA-FS (1,080,457 pairs)
`external_vqa.parquet`	Candidate-policy annotations for the external-image dataset with short answers (1,887,222 pairs)
`external_vqa_fs.parquet`	Candidate-policy external-image variant with full-sentence answers (1,887,222 pairs)

All files contain one row per question-answer pair and no image payloads.

Dataset Structure

Each parquet file contains the following columns:

Column	Description
`image_id`	Identifier of the source image within its variant
`question_id`	Unique question identifier
`answer_id`	Answer identifier
`question`	Natural language question
`answers`	Ground-truth answer
`category`	Controlled semantic question category

Citation

If you use this dataset, please cite:

@article{beliy2026brainitvqa,
  title={Brain-IT-VQA: From Brain Signals to Answers},
  author={Beliy, Roman and Cosarinsky, Matias and Heinimann, Oliver and Wasserman, Navve and Irani, Michal},
  year={2026}
}

Downloads last month: 79