Key highlights from Ultralytics at CVPR 2025

Abirami Vina

3 min read

June 18, 2025

Join us as we recap Ultralytics' experience at CVPR 2025, from booth highlights to real-world computer vision use cases and memorable moments in Nashville.

Last week, from June 11th to 15th, the Ultralytics team was on the ground in Nashville, Tennessee, for CVPR 2025 - the IEEE/CVF Conference on Computer Vision and Pattern Recognition. It’s one of the key annual events in the computer vision and AI research space. 

Held at the Music City Center, this year’s conference brought together over 12,000 attendees from around the world. From academic researchers and engineers to industry leaders and developers, the event was packed with people pushing the boundaries of Vision AI. 

In particular, CVPR 2025 featured cutting-edge research and conversations around everything from deep learning in computer vision and multimodal learning to generative AI models, foundation models, and real-world applications like autonomous driving and 3D scene understanding.

Fig 1. CVPR 2025 at Nashville, Tennessee.

For Ultralytics, the week at CVPR 2025 was focused on connecting with the computer vision community, sharing demos, and learning how our Ultralytics YOLO models are being used across industries. At Booth 1512, we showcased our latest Vision AI solutions, talked with hundreds of visitors about practical use cases, and saw firsthand how our work powers innovations across sectors like robotics, retail, research, and more. 

So let’s take a look at some of our favorite moments from the event, highlighting key takeaways, conversations, and community connections. 

A look at CVPR Nashville 2025

CVPR was first introduced in 1983 and officially backed by the IEEE in 1985. Over the years, it has grown into one of the most influential conferences in the field of computer vision and pattern recognition, bringing together researchers, engineers, and innovators worldwide. Since 2012, it has been co-sponsored by the Computer Vision Foundation (CVF), a non-profit organization that supports open access to computer vision research and plays a vital role in organizing the event. 

Held annually, CVPR has become known for its technical depth and for helping shape the direction of the entire field, from early breakthroughs in object recognition to today’s advances in Vision AI and deep learning. This year’s edition saw more than 13,000 paper submissions, a 13% increase from last year. With an acceptance rate of just 22%, the conference continued its tradition of spotlighting high-impact research.

Here’s a quick glimpse at some of the major research themes that stood out at CVPR 2025:

  • 3D scene understanding and neural rendering: Researchers are exploring how to reconstruct and represent the world in three dimensions using multiple views or sensor inputs to advance applications in robotics, augmented reality, and simulation.
  • Image and video synthesis using generative AI: This area focuses on creating high-quality visuals from minimal input, like text prompts or sketches, to power next-gen tools for creative media and interactive systems.
  • Multimodal learning and vision-language models: These models combine visual and language data to improve understanding, reasoning, and communication, paving the way for more general and adaptable AI systems.

Setting up at the Music City Center for CVPR

Before the conference officially kicked off, our team arrived in Nashville to get everything in place for CVPR 2025. 

We spent the day setting up at the Music City Center, where Booth 1512 became our home base for three full days of demos, conversations, and community building. With gear in tow, we unpacked hardware, configured live YOLO demos, and carefully arranged our setup to make sure everything would run smoothly in the fast-paced expo environment. 

The space gradually filled with people unloading gear, testing demos, and assembling their booths. Watching it all come together reminded us how global and collaborative CVPR truly is.

After a productive afternoon, the team shifted gears and headed out to explore downtown Nashville. Naturally, they made a stop on Broadway, where live country music spills out of every bar and restaurant. 

Some of the team picked up cowboy boots (a first for a few of us), and they wrapped the evening with a classic local favorite: hot chicken from Hattie B’s. It was the perfect mix of work and fun. The booth was ready, and the team was recharged to see what CVPR 2025 had in store.

Conversations about computer vision and AI research

The first day of the expo started with the Ultralytics team on-site and ready to go. Nuvola Ladi, our Digital Content Manager, Zinnia Pourdad, our Events Manager, and Francesco Mattioli, our Machine Learning Engineer, arrived at the Music City Center by 9 a.m. With final setup checks complete and demos prepped, we were ready to welcome attendees as soon as the doors opened.

By 10 a.m., the crowd began pouring in, and the energy on the expo floor picked up immediately. Our booth quickly filled with visitors, and our Ultralytics merchandise, especially the tote bags and stickers, was gone within minutes. It was a fun icebreaker that helped start conversations and draw people in.

Fig 2. Nuvola Ladi, Zinnia Pourdad, and Francesco Mattioli at Booth 1512.

​​Throughout the day, we had the chance to talk to researchers, engineers, and developers working on a wide range of projects. There was a lot of interest in our live YOLO demos, and we heard about use cases across robotics, healthcare, manufacturing, retail, and beyond. It was inspiring to see how widely our models are being used, from early academic research to deployed enterprise systems.

We also got to reconnect with collaborators in the AI space, including teams from Intel OpenVINO and Sony Semiconductor Solutions. Sharing the floor with other innovators helped reinforce just how collaborative and fast-moving the Vision AI field continues to be.

Toward the end of the day, the team headed to Printers Alley to enjoy some live country music and unwind. After a packed day of conversations and incredible energy on the floor, it was the perfect way to close out our first day at CVPR 2025.

Fig 3. The team had the chance to explore Printers Alley.

Day 2 at the computer vision conference in Nashville

Day two of CVPR 2025 picked up right where the first day left off. Booth 1512 stayed lively throughout the day, with returning visitors and new attendees stopping by to learn more about our Ultralytics YOLO models and share their own work.

For instance, one of the day’s most memorable moments was when Intel Robodog visited. Seeing advanced hardware paired with practical vision models in action really captured the spirit of the event.

Fig 4. Intel Robodog visited the team at Booth 1512.

Later in the evening, our team attended the After.CVPR() Developer Meetup, hosted by Intel. It was a relaxed and collaborative space filled with engineers, researchers, and developers from across the AI ecosystem. 

The meetup featured live demos and meaningful conversations that carried on well into the night. During the event, Francesco Mattioli presented a joint demo featuring Ultralytics and OpenVINO, showcasing how our models can be integrated into efficient inference pipelines. 

He walked through how models like Ultralytics YOLO11 can be exported to the OpenVINO format to achieve faster, optimized inference on Intel hardware. The demo highlighted support for central processing units (CPUs), graphics processing units (GPUs), and neural processing units (NPUs). It drew substantial interest from teams working on edge AI and real-time applications.

Fig 5. A glimpse of the After.CVPR() Developer Meetup.

Deep learning in computer vision applications at CVPR

Even on the final day of CVPR 2025, the Music City Center was buzzing with activity. Attendees were still exploring booths, having thoughtful conversations, and making the most of their time at the conference.

One of the highlights of the day was checking out the poster sessions - a core part of how research is shared at CVPR. Unlike traditional stage presentations, poster sessions are more informal and interactive. 

Each researcher sets up a large printed poster summarizing their work and stands nearby to explain it, answer questions, and discuss their findings with anyone who stops by. It’s a great way to learn about a wide range of topics in a short amount of time, and it creates space for real conversations between attendees and presenters.

We saw a wide variety of posters covering everything from vision-language models and 3D scene understanding to real-world applications of AI in robotics and healthcare. The sessions gave us a closer look at the latest ideas being explored across the field and a chance to connect directly with the people behind them.

With the event coming to an end, we left feeling more motivated than ever. The week was packed with learning, collaboration, and real momentum in Vision AI, and we’re excited to take those insights forward into what’s next.

Key takeaways

CVPR 2025 was a great week of learning, sharing, and connecting. From cutting-edge research to hands-on demos and great conversations, the conference showed just how quickly Vision AI is moving and how impactful and supportive the community behind it really is.

It was inspiring to see Ultralytics YOLO models in action across so many real-world and research applications. More than anything, CVPR 2025 made it clear that computer vision is at the heart of AI innovation, with so much more still to come.

Join our community and dive into Vision AI on our GitHub repository. Learn about innovations like AI in retail and computer vision in manufacturing on our solutions pages. Check out our licensing plans and begin your AI journey today!

Let’s build the future
of AI together!

Begin your journey with the future of machine learning

Start for free
Link copied to clipboard