MultiNFT Collection Banner

About

MultiNFT is a comprehensive multimodal dataset of non-fungible tokens (NFTs), sourced from OpenSea. To ensure data accuracy and consistency, the dataset is constructed using official APIs and adopts a three-stream data acquisition architecture that integrates sales records, asset metadata, and standardized image processing.


Spanning a four-year period from 1 May 2021 to 30 April 2025, MultiNFT captures 2.38M transactions involving over 500K unique NFT assets. By encompassing multiple market cycles, the dataset offers a rich foundation for digital asset analytics (e.g., price determinants and liquidity trends), market prediction (e.g., temporal movements and macro event impacts), and multimodal learning (e.g., combining image, metadata, and transaction data for classification or generative tasks).

2.38M Transactions
500K NFT Assets
50 PFP Collections
4 Years Time Span

Sample Gallery

Data Description

Hierarchical Multimodal Data Structure
1
Collections
collections.csv
50 PFP Collections • Collection Metadata
Foundation layer containing basic information for 50 carefully selected PFP collections from OpenSea's top trading volume, including metadata, contracts, and collection-level statistics.
2
NFT Assets
assets.csv
523,020 Individual NFTs • Asset Metadata
Individual NFT metadata for all assets across collections, each linked to its parent collection via collection field.
🖼️ Images
images/{collection}/{id}.png
224×224 RGB
Sample NFT
boredapeyachtclub/0.png
🏷️ Attributes
attributes/{collection}_attributes.json
Trait data
"0": [
  {
    "trait_type": "Earring",
    "value": "Silver Hoop"
  },
  {
    "trait_type": "Background", 
    "value": "Orange"
  },
  {
    "trait_type": "Fur",
    "value": "Robot"
  }
]
3
Transactions
sales_raw.csv / sales_cleaned.csv
2.38M Transactions • 4 Years (2021-2025)
Complete transaction records linked to assets via (nft_collection, nft_identifier). Cleaned version includes USD conversion and filtered payment tokens.
collection
Collections ↔ Assets
(collection, identifier)
Assets ↔ Sales
(collection, identifier)
Assets ↔ Images/Attributes

50 PFP Collections

collections.csv contains basic information for 50 carefully selected PFP collections from OpenSea's top 100 trading volume collections, all created between April 2021 and March 2023. NSFW assets are excluded from the dataset to ensure data quality and compliance.

Collections Table

Variable Name Data Type Description
collection string Collection slug. A unique string to identify a collection on OpenSea
name string Name of the collection
description string Description of the collection
image_url string Square image used to represent the collection
banner_image_url string Banner image used to represent the collection
owner string The public blockchain address of the owner
safelist_status object Status of the collection verification requests. Options: not_requested, requested, approved, verified, disabled_top_trending
category string Category of the collection (e.g. PFPs, Memberships, Art)
is_disabled boolean If the collection is currently able to be bought or sold using OpenSea
is_nsfw boolean If the collection is currently classified as 'Not Safe for Work' as defined in OpenSea's NSFW Policy
trait_offers_enabled boolean If trait offers are currently being accepted for the collection
collection_offers_enabled boolean If collection offers are currently being accepted for the collection
opensea_url string OpenSea Link to collection
project_url string External URL for the collection's website
wiki_url string External URL for the collection's wiki
discord_url string External URL for the collection's Discord server
telegram_url string External URL for the collection's Telegram group
twitter_username string Username for the collection's Twitter account
instagram_username string Username for the collection's Instagram account
contracts array Contracts for the collection
editors array List of editor addresses for the collection
fees array List of fees for the collection including creator earnings and OpenSea fees
required_zone string If defined, the required zone for all orders for this collection
rarity object Rarity data for the Collection
payment_tokens array Payment tokens for the collection
total_supply integer The total supply of the collection (minted minus burned)
created_date date The date the collection was created

Assets

assets.csv includes detailed metadata for 523,020 NFT assets from these 50 collections, with all assets having valid image and attribute metadata.

NFT Assets Table

Variable Name Data Type Description
identifier string The NFT's unique identifier within the smart contract (also referred to as token_id)
collection string Collection slug. A unique string to identify a collection on OpenSea
contract string The public blockchain address of the contract
token_standard string ERC standard of the token (erc721, erc1155)
name string Name of the NFT
description string Description of the NFT
image_url string Link to the image associated with the NFT
display_image_url string Link to the image that is display on the OpenSea website
display_animation_url string Link to the animation that is display on the OpenSea website
metadata_url string Link to the offchain metadata store
opensea_url string Link to the NFT on OpenSea
updated_at string Last time that the NFT's metadata was updated by OpenSea
is_disabled boolean If the item is currently able to be bought or sold using OpenSea
is_nsfw boolean If the item is currently classified as 'Not Safe for Work' by OpenSea as defined in OpenSea's NSFW Policy

Sales

Sales data spans four years (May 1, 2021 - April 30, 2025) across all 50 collections. sales_raw.csv contains comprehensive transaction records excluding NSFW/disabled assets and zero-price transactions, with buyer/seller addresses encrypted for privacy protection.

Raw Sales Table

Variable Name Data Type Description
event_type string Defaults to 'sale'
order_hash string Order hash for the order which was fulfilled
chain string OpenSea supported chains
protocol_address string Exchange contract address which fulfilled the order
closing_date datetime Posix timestamp converted to datetime
nft_identifier string The NFT's unique identifier (token_id)
nft_collection string Collection slug
nft_contract string Public blockchain address of the contract
nft_token_standard string ERC standard (erc721, erc1155)
nft_name string Name of the NFT
nft_description string Description of the NFT
nft_image_url string Link to the image
nft_display_image_url string Link to displayed image
nft_display_animation_url string Link to displayed animation
nft_metadata_url string Link to offchain metadata
nft_opensea_url string Link to NFT on OpenSea
nft_updated_at string Last metadata update time
nft_is_disabled boolean If item can be bought/sold
nft_is_nsfw boolean If classified as NSFW
quantity integer Number of assets transferred
seller string Seller of the NFT
buyer string Buyer of the NFT
payment_quantity string Amount of tokens in the order
payment_token_address string Contract address for ERC20 token
payment_decimals integer Number of decimals the token uses
payment_symbol string Symbol of the token (ETH, WETH, etc.)
transaction string Transaction hash
event_timestamp datetime Posix timestamp converted to datetime

sales_cleaned.csv filters transactions to four major payment tokens (ETH, WETH, DAI, USDC), removing 97 transactions with other tokens, and integrates Yahoo Finance daily close prices to calculate USD values and logarithmic prices for enhanced price analysis.

Cleaned Sales Table (Added Variables)

Variable Name Data Type Description
payment_price float Actual payment amount (quantity ÷ 10^decimals)
payment_usd_price float Payment amount converted to USD
log_usd_price float Natural logarithm of USD price

Download

Dataset hosted on Zenodo. Click individual files to download directly, or visit the complete dataset page for detailed information and citations.

Collections & Assets

Basic information for 50 PFP collections and metadata for 523,020 NFT assets.

Click to download:

Sales

Four years of transaction records with raw and cleaned versions for analysis.

Click to download:

Images & Attributes

523,020 standardized images (224×224 PNG) and corresponding trait attribute data.

Click to download:

Complete Dataset

Full multimodal dataset package with all sales, metadata, and image data for research.

Access Complete Dataset

Contact

Shuying Liu CORRESPONDING
2595459L@student.gla.ac.uk
University of Glasgow
Adam Smith Business School
Bowei Chen CORRESPONDING
bowei.chen@glasgow.ac.uk
University of Glasgow
Adam Smith Business School
Cathy Yi-Hsuan Chen
CathyYi-Hsuan.Chen@glasgow.ac.uk
University of Glasgow
Adam Smith Business School
Southwest University of Finance and Economics
School of Finance, SWUFE
University College London
Department of Computer Science