▲ Emotionwave Dataset Construction and Processing Process Diagram and Artificial Intelligence Model YOURMT3+ Leading AI-based Media Technology Company…
▲ Emotionwave Dataset Construction and Processing Process Diagram and Artificial Intelligence Model YOURMT3+
Emotionwave, a leading AI-based media technology company (CEO Rio Jang), unveiled the process and results of valuing AI music data, along with the latest AI technologies and processes for data generation and processing, at the 2024 Data Promotion Week held at the Fairmont Ambassador Seoul Grand Ballroom in Yeouido on December 10th. This presentation was conducted as a case study of data value evaluation, and CEO Rio Jang gave the presentation.
During the presentation, CEO Rio Jang explained the systematic generation process of AI music data built by Emotionwave and the impact this data has had on service and product innovation. In particular, he emphasized the importance of artificial intelligence technology for data accuracy and refinement, and presented specific ways to enhance the competitiveness of technology companies through this.
Emotionwave announced that it introduced Cross-Dataset Stem Augmentation and Intra-Stem Augmentation technologies to improve data accuracy. These technologies enable the integrated use of various music datasets and effectively solve problems of data imbalance and insufficient annotation. Furthermore, by implementing Real-Time Data Augmentation techniques such as Pitch-Shifting of audio sources, they realized diverse data scenarios, further enhancing the universality and accuracy of the AI model.
In particular, the YourMT3+ technology developed by Director Jang Sung-kyun (CMTO, Chief Music Technology Officer) of Emotionwave was applied. This technology, which utilized Spectral Cross Attention(SCA) and Mixture of Experts(MoE) for encoder upgrades and a multi-channel decoder (Multi-T5), significantly improved the analysis accuracy and efficiency of data. By converting PCM data into single or multi-track scores, it efficiently processed complex music data and greatly enhanced data quality.
This data was utilized as a dataset for our brand's music generation models, such as RIMA, MUTA, and StageCloud, and was also applied to the processing process for creating the dataset. Through this, Emotionwave established its own artificial intelligence data and core data processing processes for commercialization, gaining recognition for its technological value in the industry.
In the presentation, CEO Rio Jang stated, "The core competency that allows us as a technology company to look further into the future, whether it's 5 or 10 years, lies in data," and added, "I want to emphasize that the beginning and end of information technology is data." He further added, "We are pleased to have gained credibility for our technology by having our data value and technology recognized by a specialized institution accredited by the Ministry of Science and ICT."
Through this technology unveiling and data valuation announcement, Emotionwave once again solidified its position as a leading company driving data-based innovation by presenting new possibilities that AI and data can create in the music industry.