Dolphin's new release deepens its understanding of Vision AI datasets, central to a data-centric approach to MLOps.
MindTech has announced the immediate availability of its latest release of its Dolphin platform, which performs AI-based training data analysis, visualization and curation of image datasets. Computer vision.
Featured article: The Role of AI in Cybersecurity: Protecting Digital Assets from Cybercrime
The new release empowers data scientists to gain a deeper understanding of their training datasets through custom network fit analysis and full annotation visualization. ML Engineer A deep understanding of the training dataset through appearance and content analysis is critical to achieving more accurate and robust learning. AI system.
The new Custom Network Fit Analysis feature allows data scientists to specify custom classification networks, such as Yolo, to determine exactly how the network sees a particular image (whether real or synthetic) when compared to other images in the analyzed dataset. Dolphin examines the internal embeddings of the network, which are typically hidden inside a “black box,” and automatically identifies the most relevant values from hundreds of values per image through Principal Component Analysis (PCA), producing easy-to-interpret, human-readable 2D charts.
This analysis helps engineers quickly understand shortcomings in the dataset, such as sparsity or diversity issues, which are important to identify and resolve to ensure the network is robust when deployed in the real world. A second key area is data drift, which is when the original training data is no longer relevant or new training data is needed due to changing real-world conditions. Data drift can be identified without the need to label new data.
read more: Deloitte teams up with HPE and NVIDIA to develop generative AI solutions
The second key enhancement is annotation visualization from within the platform, allowing engineers to quickly review selected images, such as outliers, and investigate and validate associated annotations. Dolphin is fully containerized, making it easy to install on-premise or in the cloud, and easily accessible from any supported browser. Dolphin is part of the Mindtech DataOps platform, and the actionable intelligence created by Dolphin can be used by Chameleon, Mindtech's synthetic data creation platform, to easily and quickly create synthetic data to broaden the scope of training data and improve the performance of AI models.
Chris Longstaff, vice president of product management at MindTech, said: “We are living in an era where AI-based systems are facing greater scrutiny than ever before due to societal and government pressures, so understanding how robust they are is more important than ever. With the introduction of legislation such as the EU AI Act and the Colorado Artificial Intelligence Act, the onus is on technology providers to demonstrate they have full data governance and transparency. MindTech's Dolphin platform can be a key part of meeting these requirements, helping to identify bias, diversity and sparsity in data.”
Mindtech will be at stand 2039 at the CVPR Expo in Seattle from June 19-21 to demonstrate the platform.
Security and AI features: Employee data risks in an AI-driven world
[To share your insights with us as part of editorial or sponsored content, please write to psen@itechseries.com]
