Machine learning to generate insights in the digital creative production process

• Grant number: 22/14538-1
• Support Opportunities: Research Grants - Innovative Research in Small Business - PIPE
• Start date: February 01, 2024
• End date: July 31, 2025
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computer Systems
• Agreement: MCTI/MC
• Principal Investigator: Adalberto Generoso da Costa
• Grantee: Adalberto Generoso da Costa
• Company: Yapoli Desenvolvimento de Programas Ltda
• CNAE: Desenvolvimento e licenciamento de programas de computador customizáveis ; Desenvolvimento e licenciamento de programas de computador não-customizáveis ; Consultoria em tecnologia da informação
• City: São Paulo
• Pesquisadores principais: Felipe Francesco Pereira Lopes da Costa ; Rodrigo Freitas Lima
• Associated researchers: Jéssika da Silva

Abstract

Founded in 2018, the startup Yapoli has large customers operating in various business segments. The startup offers the market a digital asset management platform called DAM - Digital Asset Management, which stores any digital artifacts of its customers in the cloud and makes indexing and all their lifecycle management simple. Currently, the platform manages a volume of approximately 500 thousand assets and 5 terabytes of assets. Artifacts of the image or video type are subjected to a process of extracting information through computer vision artificial intelligence algorithms provided by cloud computing companies such as the Amazon Rekognition (AWS) and Vision AI (GCP) service. Such tools are already used by Yapoli for extraction and indexing, however these platforms use generic training algorithms and, therefore, extract very basic information, insufficient to offer relevant insights to the business. The technological development opportunity identified with customers is to generate information regarding the result that digital advertising artifacts generate. To this end, it is understood that some digital assets can influence events in the results of customer marketing campaigns, ranging from conversion in sales and retention on website page to interaction with digital assets on social networks such as views, shares, likes and comments. Currently, companies' marketing campaigns rely exclusively on assumptions, which are primarily based on the experience of those involved, without a systematic analysis of data to support the decision to create advertising campaigns. Given this context, the present project aims to evolve in the development of a system that is capable of extracting qualified information from digital assets through computer vision models, capturing events from digital channels that relate to such assets and performing the crossing of data to obtain insights that are relevant to advertising campaigns. (AU)