Green check
Link copied to clipboard

YOLO VISION 2022: The New Frontier of Vision AI

YOLO VISION has brought AI experts and enthusiasts alike together.

Facebook logoTwitter logoLinkedIn logoCopy-link symbol

Our first-ever YOLO VISION took place on September 27th, 2022. From the entrance of AI in the automotive industry to real-time analysis of fruit production, we listened to inspiring talks from YOLOv5 users across the board.

Something that made this event special was the wide variety of backgrounds of the speakers. Joining representatives from 18 participating companies, the speakers delivered insights from every aspect of the ML process. Among them, are our partner companies such as Comet, Deci, ClearML, Paperspace, and Roboflow, as well as others in the open-source space like Chinese giants Baidu, Meituan, and OpenMMLabs.

Redefining State-Of-The-Art With YOLOv5

Wondering about the story behind the creation of YOLOv5 and the methodology used for R&D?

Dive into the details of the holistic approach used to choose the best architectures with Glenn Jocher, our Founder & CEO here at Ultralytics, and Ayush Chaurasia, our ML Engineer.

Great model architectures like YOLOv5 are crucial to getting useful results in machine learning. But models are only as good as their datasets. Joseph Nelson, CEO & Co-founder at our partner Roboflow, showed the impact of dataset quality on production results. The insights are informed by over 10,000 vision training jobs and Roboflow Universe’s open-source community of 90,000+ datasets.

In his session, Joseph also showcased key differences in research vs production that enable developers to hack their datasets to get meaningful results faster.

Learn about dataset quality and its impact on getting your CV model to production value!

Best Practices for Validating your ML Model & Data Before Deployment

Every piece of traditional software today goes through comprehensive tests of various types before deployment, significantly reducing the risk of production faults.

How can we adapt these ideas to the statistically-oriented world of ML?

Aishwarya Srinivasan, Data Scientist at Google & Open Source Developer Advocate at Deepchecks, talks about the mere excitement behind building solutions that are able to solve real-world challenges. At Google, she builds machine learning solutions for customer use cases, leveraging core Google products including TensorFlow, DataFlow, and AI Platform.

Aishwarya joined us at YOLO VISION to discuss best practices and practical tips for extensively testing and analyzing your model. Check out her talk to learn the difference between Testing Software and Testing ML.

Open Source Projects Enabling the Future of Computer Vision AI

We hosted a groundbreaking panel where we gathered other members of the YOLO architecture family together as well as other top open-source vision AI architectures in the space.

Here, Meituan’s YOLOv6, OpenMMLab CN's MMDetection, and Baidu, Inc.'s PaddlePaddle joined us as Ultralytics' YOLOv5 to discuss open-source projects enabling the future of vision AI.

This was the first time ever, that these top vision AI repositories have shared the stage. If you missed this panel, watch this video where Bo Zhang, Glenn Jocher, Guanzhong Wang, Wenwei Zhang, and Yixin Shi discussed their choice of frameworks, designs, the evolution of repository structure, and more!

As our CEO Glenn Jocher says, “We all got to learn from each other’s tools and experiences.”

Visual Data is Exploding

Visual data management systems are lacking in all aspects: storage, quality, search, analytics, and visualization. As a consequence, companies, and researchers are losing product reliability, working hours, wasted storage, computing, and most importantly, the ability to unlock the full potential of their data.

In this talk, Dr. Danny Bickson taught us how to solve this problem with his popular free GitHub tool, Fastdup.

FastDup is a tool for gaining insights from a large image collection. It can find anomalies, duplicate and near duplicate images, clusters of similarity, and learn the normal behavior and temporal interactions between images. It can be used for smart subsampling of a higher-quality dataset, outlier removal, and novelty detection of new information to be sent for tagging.

An expert in big data analytics and large-scale machine learning, Danny Bickson has more than 15 years of experience in the high-tech industry. You might know him from Turi, a machine learning platform that creates big data analytics products for its users. In 2016, Turi was acquired by Apple where Dr. Danny Bickson worked as Senior Data Science Manager for several years.

Your Doorway To Vision AI

And finally, it was our pleasure to formally announce the launch of our Ultralytics HUB!

Ultralytics HUB is our no-code solution to train and deploy AI models in three easy steps! Bring your models to life by choosing what data for it to learn from.

Our experts, and creators of the tools, Kalen Michael and Sergio Sánchez, took us on a walkthrough of Ultralytics HUB and explained all the features and functionalities Learn more about Ultralytics HUB and start creating your models for free!

Find all recorded sessions on our YouTube channel!

We are thrilled with the turnout for YOLO VISION and happy to create an event where experts from around the world can join to learn about vision AI, stay up to date with us by following us on social media. See you next year at YOLO VISION 2023!

Let’s build the future
of AI together!

Begin your journey with the future of machine learning

Read more in this category