Top AI & Computer Vision Conferences to Attend in 2026 (Plus a 2027 Sneak Peek 👀)

For business leaders, data engineers, and researchers, attending an AI or computer vision conference should be a top priority.

These gatherings showcase breakthroughs in generative AI models, deep learning, and annotation, but the biggest value of attending a computer vision or AI conference lies in their ability to connect you with industry leaders. Through expert sessions, workshops, and hands-on demonstrations, attendees get the chance to learn AI strategies from the top thought leaders in the space.

The 2026 calendar is unusually dense in the back half of the year, and the events have specialised. Rather than one general "AI conference" circuit, there are now distinct tracks for robotics, medical imaging, industrial deployment, and infrastructure, and the right choice depends far more on what you build than on which event is largest. Below, we break them down by discipline, with dates, costs, and virtual access options so you can plan a realistic year.

The Remaining 2026 AI Conferences at a Glance

We will go over each conference in more detail later in the article. But if you want a quick overview of the 10 biggest conferences, the table below will help.

#1 Computer Vision Conferences

ECCV (European Conference on Computer Vision)

ECCV is Europe’s flagship computer vision conference, alternating years with ICCV. It emphasizes both foundational theory and practical applications in areas such as 3D reconstruction, remote sensing, video understanding, and multimodal models.

The 2026 edition splits across two venues in Malmö: workshops and tutorials open the week, followed by the main single-track programme.

• Date: September 8-12, 2026. Workshops and tutorials run September 8-9; the main conference runs September 10-12.
• Location: Malmö Arena and Malmömässan, Malmö, Sweden.
• Virtual Access: Yes. A virtual pass costs $175 ($116 for students) and includes access to sessions on a dedicated online platform. Note that accepted papers must be covered by a full in-person registration, virtual and student passes do not count.
• Cost: Full registration is $805 early-bird, rising to $1,045; student registration is $440, rising to $557. Workshop and tutorial day passes run $125-$250.

Focus / Who Should Attend:

ECCV is ideal for European researchers, startups, and business leaders looking to collaborate on projects that link vision with AI-driven business value. It offers a strong mix of academic sessions, poster presentations, and networking opportunities.

If you are choosing just one academic vision conference in the back half of 2026, this is the one with the broadest industry presence, the exhibition floor draws the major annotation, hardware, and foundation-model vendors alongside the research programme.

BMVC (British Machine Vision Conference)

Now in its 37th year, BMVC is the UK’s principal computer vision conference, run by the British Machine Vision Association. It is deliberately smaller and considerably cheaper than CVPR or ECCV, with a single-track programme that makes it far easier to actually meet the people whose work you follow.

For teams without the budget to send people to Malmö or Denver, BMVC is the most cost-effective way to stay current with serious vision research, and it is a notably friendly venue for first-time presenters and PhD students.

• Date: November 23-26, 2026.
• Location: Lancaster Town Hall, hosted by Lancaster University, UK.
• Virtual Access: Not announced for 2026. Previous editions have published proceedings openly after the event.
• Key Deadlines (2026 cycle, now closed): Abstracts were due May 22, papers May 29, with author notifications on August 7 and camera-ready papers due August 28. Expect a similar cadence for 2027.

Focus / Who Should Attend:

BMVC suits PhD students, academic researchers, and UK or European engineering teams who want depth without the scale and cost of the larger conferences.

The BMVA also runs one-day symposia throughout the year on focused themes, AI for Visual Arts on December 2, 2026 is the next of these, which are worth watching if you want a low-commitment way to engage with the UK vision community.

#2 Robotics and Physical AI Conferences

IROS (International Conference on Intelligent Robots and Systems)

IROS is one of the two dominant robotics conferences worldwide, run jointly by the IEEE Robotics and Automation Society and the Robotics Society of Japan. Its centre of gravity is robot perception, learning, control, and human-robot interaction, which makes it the most directly relevant event of 2026 for anyone building the vision stack inside a physical system.

The scale is substantial and growing: IROS 2025 drew 4,306 paper submissions and accepted 1,991 of them, an acceptance rate of 46.2%, up from 3,400 submissions the year before.

The 2026 host city is a deliberate choice. Pittsburgh has been nicknamed “Roboburgh” off the back of Carnegie Mellon’s Robotics Institute and the autonomy companies that spun out of it, so the local industry presence should be unusually strong.

• Date: September 27 - October 1, 2026.
• Location: David L. Lawrence Convention Center, Pittsburgh, Pennsylvania, USA.
• Virtual Access: Not announced. IROS has historically been an in-person event with proceedings published through IEEE Xplore.
• Cost: Registration rates were not published at the time of writing. IEEE Life Members attend free.

Focus / Who Should Attend:

IROS is for robotics engineers, perception leads, and research teams working on autonomy, manipulation, drones, and embodied AI. If your annotation work involves multi-sensor data, LiDAR point clouds, stereo rigs, or synchronized camera arrays, this is where the state of the art gets argued about.

One scheduling note: IROS now runs on exactly the same dates as MICCAI. If your team spans robotics and medical imaging, you will have to split up.

#3 Medical Imaging and Healthcare AI Conferences

Clinical imaging is the most regulated corner of computer vision, and it has its own conference circuit, its own benchmarks, and its own standards of evidence.

MICCAI (Medical Image Computing and Computer Assisted Intervention)

MICCAI is the premier global gathering at the intersection of medical imaging, machine learning, and clinical practice, now in its 29th edition. It is where segmentation, registration, and computer-assisted intervention research gets presented, and increasingly, where the conversation turns to whether any of it survives contact with a hospital.

Important change for 2026: the conference has been relocated. It was originally scheduled for Abu Dhabi on October 4-8, but the MICCAI Board moved it to Strasbourg, France, citing ongoing uncertainty in the Middle East. The dates shifted with it. If you booked travel off an earlier version of the schedule, this is the one to double-check.

The programme is broader than a straight paper track: alongside keynotes and workshops it includes ClinicCAI (a dedicated clinical day), open data sessions, public challenges, and a Startup Village.

• Date: September 27 - October 1, 2026 (moved from the originally announced October 4-8).
• Location: Strasbourg Convention Center, Strasbourg, France (relocated from Abu Dhabi, UAE).
• Virtual Access: Not confirmed following the venue change.
• Scale: MICCAI 2025 accepted 1,014 papers at an acceptance rate of roughly 29%, in line with recent years.

Focus / Who Should Attend:

MICCAI is essential for teams building AI for radiology, pathology, surgical guidance, or any clinical imaging application. It is also the best single place to understand how regulated annotation actually works, clinician-in-the-loop labelling, inter-rater agreement, and the evidence standards that medical AI has to clear.

The public challenges are worth attending even if you are not presenting: they are effectively an annual benchmark of what is solved and what is not in medical image analysis.

#4 Industrial and Embedded Vision Conferences

Where computer vision meets a production line or a shipping product. These events are less about novel architectures and more about what actually survives deployment.

VISION 2026

VISION in Stuttgart is not an academic conference, it is the world’s leading trade fair for machine vision, and it fills a gap the research circuit does not cover. This is where you go to actually specify and buy a vision system: cameras, sensors, optics, lighting, frame grabbers, and the software that ties them together.

The most recent edition drew 483 exhibitors from 38 countries and around 8,000 visitors. It runs biennially, so 2026 is an on-year, the next one after this is 2028.

The accompanying Industrial Vision Days lecture forum runs alongside the exhibition and is free to attend with a fair ticket, covering embedded vision, hyperspectral imaging, and deep learning in industrial settings.

• Date: October 6-8, 2026.
• Location: Messe Stuttgart, Stuttgart, Germany.
• Virtual Access: This is an in-person trade fair.
• Cost: A VISION ticket also grants entry to the co-located Motek, EFX, Quantum Effects, and interbad exhibitions running at the same venue.

Focus / Who Should Attend:

VISION is for product engineers, systems integrators, procurement teams, and anyone deploying computer vision on a factory floor or production line. If your problem is defect detection, bin picking, optical inspection, or robotic guidance, three days here will teach you more about the practical constraints than any research conference.

It is also the clearest window into what industrial vision hardware will look like over the next two years, which matters if you are designing an annotation and training pipeline around sensors that do not exist yet.

Edge AI and Vision Innovation Forum

The Innovation Forum is a focused, single-day event run by the Edge AI and Vision Alliance, the same organization behind the much larger Embedded Vision Summit. It concentrates on the practical business of getting visual AI into shipping products: silicon selection, model optimization, power budgets, and deployment economics.

One important caveat: unlike everything else on this list, the Innovation Forum is invitation-only and open exclusively to Alliance member companies. If your organization is not a member, the Embedded Vision Summit in May is the equivalent open event.

• Date: September 2026, San Jose. Confirm the exact date directly with the Alliance, published listings currently disagree between September 10 and September 16.
• Location: San Jose, California, USA.
• Virtual Access: In-person, and member-only.

Focus / Who Should Attend:

This one is for product and engineering leaders at Alliance member companies who are actively shipping edge vision hardware. If that is not you, note the date anyway: the technical themes surfaced here tend to set the agenda for the following year’s Embedded Vision Summit.

#5 Machine Learning Tooling and Infrastructure Conferences

PyTorch Conference North America

The PyTorch Conference is the annual gathering for the framework that most computer vision work is actually built on. Run by the PyTorch Foundation under the Linux Foundation, it has broadened well past PyTorch itself, the 2026 programme spans vLLM, DeepSpeed, Ray, Helion, and Safetensors, covering the full stack from kernel engineering up to agent frameworks.

Where the academic conferences tell you what is newly possible, this one tells you how to make it run at a sane cost. Expect deep technical sessions on training throughput, inference optimization, and serving infrastructure.

• Date: October 20-21, 2026.
• Location: San Jose Convention Center, San Jose, California, USA.
• Virtual Access: The 2026 edition is in-person only. Sessions are typically published to YouTube afterwards.
• Cost: $375 early-bird, $475 standard, and $925 for late registration, which makes it one of the better-value technical conferences on this list if you book ahead.

Focus / Who Should Attend:

This is the most directly useful event of the year for ML engineers, MLOps teams, and infrastructure leads. Less strategy, more of the specific knowledge that shortens a training run or halves an inference bill.

It is also the right place to bring integration questions. If you are wiring an annotation platform into a training pipeline, the people who maintain the relevant libraries are in the room.

NeurIPS (Neural Information Processing Systems)

NeurIPS is the largest and most influential machine learning conference in the world, and its fortieth edition is doing something genuinely new: running across three cities at once.

The reason is straightforward. NeurIPS 2025 in San Diego received 21,575 paper submissions, accepted 5,290 of them, and drew well over 20,000 attendees, the largest edition ever held, and past the point where any single venue comfortably works. Splitting across three continents also cuts travel distance and eases the visa problems that have kept researchers away in recent years.

Sydney is the main venue, with Atlanta and Paris running as full satellites, not overflow rooms. Each has its own programme chairs, keynotes, oral and poster sessions, and workshops. Of the 102 workshops accepted for 2026, 48 run in Sydney, 28 in Paris, and 26 in Atlanta, so the workshop programme is genuinely different at each site.

• Date: Sydney, December 6-12, 2026 (main venue). Atlanta, December 9-13, 2026. Paris, December 9-13, 2026.
• Location: Sydney, Australia; Atlanta, Georgia, USA; and Paris, France.
• Virtual Access: NeurIPS has run virtual components in recent years; confirm the 2026 arrangement per venue.
• Key Date: Paper author notifications went out September 24, 2026.

Focus / Who Should Attend:

NeurIPS is for machine learning researchers and engineers working at the frontier, foundation models, multimodal learning, generative methods, and the theory underneath them. For a computer vision team, it is where vision stops being a separate discipline and becomes one modality among several.

Practical advice: pick your venue by workshop programme rather than geography. Check which of the 102 workshops are running where before you book, because the site with the sessions you care about may not be the closest one.

#6 Enterprise Strategy and Applied AI Conferences

Broad, business-focused events aimed at leaders making buying and roadmap decisions rather than engineers making architectural ones.

NVIDIA GTC Berlin

GTC Berlin is the European edition of NVIDIA’s GPU Technology Conference, and it has grown into the single largest gathering of the European AI ecosystem. The format mirrors the San Jose event, a three-day developer conference on accelerated computing spanning training infrastructure, inference optimization, robotics, and autonomous systems.

The headline keynote from NVIDIA founder and CEO Jensen Huang runs on the middle day at the Tempodrom, a short distance from the main venue. If you follow the sector at all, that keynote is usually where the year’s hardware and platform roadmap is set out.

• Date: October 20-22, 2026. The Jensen Huang keynote is on Wednesday, October 21.
• Location: STATION-Berlin, Berlin, Germany, with the keynote at the nearby Tempodrom.
• Virtual Access: NVIDIA has historically streamed GTC keynotes and published session recordings afterwards.

Focus / Who Should Attend:

GTC Berlin suits European developers, infrastructure leads, enterprise architects, and researchers who cannot easily justify the trip to San Jose in March. The computer vision and robotics content is substantial, and the exhibition floor is the most efficient way to survey the European AI vendor landscape in one go.

Note the date clash with the PyTorch Conference in San Jose, which runs October 20-21. If you have to choose, GTC leans toward hardware, platform, and strategy; PyTorch leans toward the software you write on top of it.

Which Conference Is Right For You?

There is no single best AI conference, only the one that fits what you are trying to do. Use this as a shortcut before reading the details below.

A Quick Look At the 2027 Calendar

Several of the major venues have already confirmed dates and cities for 2027. If you are planning travel budgets a year out, these are firm enough to hold.

Two of these are worth flagging now. ICCV returns to Hong Kong in October 2027 after skipping 2026, and CVPR 2027 heads back to Seattle in June. Between them, they anchor the 2027 vision research calendar, and both are far enough out to build a travel budget around.

The Problem No Conference Will Solve for You

Every event on this list will show you a model that beats what you are running today. Very few will spend much time on the least glamorous finding in applied machine learning: the distance between a benchmark result and a system that works in production is usually a data problem, not an architecture one.

Sit through the MICCAI challenge sessions and you will hear it as inter-rater disagreement between clinicians. At IROS it shows up as sensor calibration and the cost of labeling point clouds. At VISION it is the defect class nobody photographed enough of. You come to realize it’s the same problem, different vocabulary, and it is the one that decides whether any of the research on display makes it into your product.

That is the part CVAT exists for. It is an open-source annotation platform built specifically for computer vision, images, video, and 3D point clouds, and used to build training data for exactly the applications these conferences are about. You can run it yourself, use the hosted version, or hand the labeling to our annotation team.