September 2024 Summaries

The September '24 AI, Machine Learning, and Computer Vision Meetup covered topics such as managing large-scale GPU infrastructure with GPUaaS solutions, simplifying data-model co-development for Industrial use cases using FiftyOne Teams, and more. Key speakers included Mischa van Kesteren from NexGen Cloud and Daniel Gural from Voxel51. Upcoming Meetups will take place on October 10th with topics such as leveraging machine learning to scale electric vehicle sales, exploring four-channel ML workflows, and identifying bird species through vector similarity search. The Computer Vision Meetups are sponsored by Voxel51, the company behind the open source FiftyOne computer vision toolset.

Tom Roussel, Machine Learning Engineer at Segments.ai, discusses their new plugin for FiftyOne, an AI/ML tool designed to refine datasets and models together. The Segments.ai plugin focuses on multi-sensor labeling, allowing users to annotate different sensors in a single interface and get consistent labels across all sensors. Key benefits of the plugin include reduced friction between data curation and annotation tasks, faster iteration cycles, and seamless integration with FiftyOne's data exploration capabilities. The plugin enables customers to request annotations directly within the FiftyOne interface, eliminating the need for context switching between tools. Resources for getting started with the Segments.ai plugin include a blog announcement with tutorial videos, GitHub repository for installation and configuration, and the option for a demo.

The Segments.ai plugin for FiftyOne provides machine learning teams with an integrated solution for multi-sensor data curation and annotation, allowing users to efficiently move between dataset exploration and high-quality labeling within a single workflow. Developed by Segments.ai ML Engineer Tom Roussel, this tool connects FiftyOne’s dataset inspection and error analysis capabilities with Segments.ai’s comprehensive multi-sensor labeling platform, which supports consistent annotations across multiple sensors such as cameras, lidar, and radar. This seamless integration minimizes the need for context switching between tools, reduces friction, and enhances the quality of datasets by enabling users to identify and annotate challenging cases directly within FiftyOne. The plugin's design caters to the iterative nature of developing visual AI systems, allowing for faster development cycles and improved dataset quality through streamlined data exploration, annotation, and quality control processes.

Voxel51's weekly digest includes news about ViewCrafter, a tool that synthesizes high-fidelity novel views of generic scenes using single or sparse images, video diffusion models, and point-based 3D representations. The digest also covers research on the use of AI in academic papers, OpenAI's potential restructuring to a more traditional for-profit model, various GitHub projects related to AI, Paul Graham's essay on Founder Mode, the controversy surrounding Reflection 70B, and a survey on comic understanding.

The September '24 Visual AI in Healthcare Meetup featured discussions on improving the interpretability of AI models in medical image analysis, using comparative computational AI in veterinary medicine, exploring NVIDIA's VISTA-3D and MedSAM-2 medical imaging models, and addressing instance imbalance in medical semantic segmentation. The event also included Q&A sessions with experts in the field and provided information on upcoming Meetups and ways to get involved.

Voxel51's latest newsletter highlights new integrations with Segments.ai and FiftyOne, offering seamless data annotation processes. Upcoming hackathons at the University of South Carolina, Northwestern University, and NYU provide opportunities for ML enthusiasts to tackle real-world challenges in AI technology. Additionally, a Data-Centric AI competition on Hugging Face offers participants a chance to curate a subset of a dataset while maintaining or improving model performance.

The September '24 AI, Machine Learning, and Computer Vision Meetup covered various topics such as reducing hallucinations in ChatGPT using knowledge graphs, the Data-Centric AI Competition on Hugging Face Spaces, and incorporating sensor data into Retrieval Augmented Generation (RAG) applications. The event also highlighted the importance of data curation in AI development. Upcoming Meetups will feature speakers discussing interpretable AI models in radiology, advancing comparative computational AI in veterinary oncology, deep-dives into NVIDIA's VISTA-3D and MedSAM-2 medical imaging models, and exploring instance imbalance in medical semantic segmentation. The Meetups aim to bring together communities of data scientists, machine learning engineers, and open source enthusiasts who want to share and expand their knowledge of AI and complementary technologies.

The meetup aimed to bring together communities of data scientists, machine learning engineers, and open source enthusiasts to share and expand their knowledge of AI and complementary technologies. The event featured presentations on various topics such as agentic RAG in 2024, accelerating multimodal RAG pipelines with NVIDIA and Open-Source Integrations, and using embeddings for robust OCR document search and cross-modal retrieval. The meetup also highlighted the importance of charitable giving, donating $200 to Heart to Heart International, an organization that ensures quality care is provided equitably in medically under-resourced communities and disaster situations. Upcoming meetups are scheduled, including a discussion on reducing hallucinations in chatGPT and similar AI systems, data-centric AI competition, and sensor data in RAG. The meetup is sponsored by Voxel51, the company behind the open-source FiftyOne computer vision toolset, which enables data science teams to improve the performance of their computer vision models.