The demand for large-scale diverse datasets is rapidly increasing due to the advancements in AI services impacting day-to-day life. However, gathering such massive datasets still remains a critical challenge in the AI service engineering pipeline, especially in the computer vision domain where labeled data is scarce. Rather than isolated data collection, crowdsourcing techniques have shown promising potential to achieve the data collection task in a time and cost-efficient manner. In the existing crowdsourcing marketplaces, the crowd works to fulfill consumer-defined requirements where in the end consumer gains the data ownership and the crowd is compensated with task-based payment. On the contrary, this work proposes a blockchain-based decentralized marketplace named Vision Sovereignty Data Marketplace (ViSDM), in which the crowd works to fulfill global requirements & holds data ownership, the consumers pay a certain data price to perform a computing task (model training/testing), the data price is distributed among the crowd in a one-to-many manner through smart contracts, thus allowing the crowd to gain profit from each consumer transaction occurring on their data. The marketplace is implemented as multiple smart contracts and is evaluated based on blockchain-transaction gas fees for the stakeholder interaction & by running scenarios-based simulations. Furthermore, discussions address the challenges included in maintaining data quality and the future milestones towards deployment.

https://dl.acm.org/doi/10.1145/3582515.3609556