Let's talk about ICCV. If you're in computer vision, you've heard the name. It's one of the two top-tier conferences, the other being CVPR. But what's it really like on the ground? Is it just a week of presentations, or is there more? I've been through the cycle—submitting papers, facing rejections, finally getting that acceptance, and navigating the conference itself. This isn't a generic overview. This is a practical guide from someone who's been in the trenches, covering what you actually need to know to participate successfully, whether you're a PhD student aiming for your first publication or a senior engineer scouting for talent and trends.
What’s Inside This Guide
What ICCV Is (And What It Isn't)
ICCV stands for the International Conference on Computer Vision. It happens every two years, in odd-numbered years, which gives it a slightly different vibe than the annual CVPR. The bi-annual schedule means the pressure and anticipation are higher. The community saves some of its most significant work for it.
Here's the thing newcomers often get wrong. They think ICCV is just about the papers in the main conference proceedings. That's the core, but it's only part of the ecosystem. The conference is a massive, week-long event comprising workshops, tutorials, challenges, and an industrial exhibition. The main conference talks are valuable, but some of the most cutting-edge ideas and niche research discussions happen in the parallel workshops. I've found more relevant technical debates in a workshop on "Robustness in Vision" than in some of the broader main conference sessions.
The location rotates globally—recent ones have been in Montreal, Seoul, and Venice. This rotation affects everything: cost, attendee demographics, and even the social events. A conference in Europe will draw a different crowd than one in Asia. You need to factor travel and visa logistics into your planning early.
How to Submit a Paper to ICCV: A Step-by-Step Walkthrough
The submission process is a marathon, not a sprint. Missing a nuance can lead to desk rejection or poor reviews. Let's break it down.
The Critical Timeline You Must Respect
Dates shift slightly each cycle, but the structure is consistent. The abstract submission deadline usually comes a few days before the full paper deadline. This isn't a placeholder. You must submit a complete abstract. I've seen papers get mismatched reviewers because a hastily written abstract didn't accurately represent the final work.
The full paper deadline is strict—no extensions. The review period lasts several months, followed by rebuttals. The rebuttal is your chance to clarify misunderstandings, not to re-argue the entire paper or add new experiments. Be factual and polite. Then, after the final decisions, comes the camera-ready deadline and copyright form submission.
Choosing the Right Submission Track
This is a subtle but crucial point. ICCV typically has a main conference track and sometimes separate tracks for specific themes or formats. The review criteria and competitiveness can vary. A common mistake is submitting a solid but incremental engineering paper to the main track, where it's judged against groundbreaking theoretical work. It might fare better in a more applied workshop. Always read the call for papers for each track carefully.
The Anatomy of a Successful ICCV Paper
Beyond technical novelty, presentation matters immensely. Reviewers are overloaded. Make their job easy.
- The Title and Abstract: These are your billboards. They must clearly state the problem, your core idea, and the key result. Avoid overly clever but vague titles.
- The Introduction: Frame the problem in the first paragraph. What's the gap? Why does it matter? Summarize your method and results by the end of the introduction. A reviewer should know your contribution without reading the whole paper.
- Figures and Visualizations: This is computer vision. Your figures must be impeccable. They should tell the story. Include clear qualitative results (visual comparisons) alongside quantitative tables.
- The Code and Data Clause: While not always mandatory for submission, the expectation for releasing code is now the norm. Plan for this from the start. A note about "code will be released upon acceptance" is standard, but having a repository ready (even if private) makes the camera-ready process smoother.
One non-consensus opinion: spending a week polishing your figures and writing is often more impactful than spending that week running one more ablation study with a 0.2% improvement. Clarity beats marginal gains when it comes to reviewer perception.
How to Navigate the ICCV Conference Week
You got in. Now what? The conference week is a whirlwind. Here’s how to structure it.
Before You Go: The Essential Pre-Conference Checklist
Don't just book flights and hotel.
- Build Your Schedule: Use the conference app or website. Flag must-see talks in your area. But also flag 2-3 talks outside your immediate focus—this is how you discover new ideas. >Plan Your Poster Session: If you have a poster, know your slot. Prepare a 1-minute, 3-minute, and 5-minute explanation. Have business cards or a QR code linking to your paper/code.
- Reach Out for Meetings: Identify 5-10 people you'd like to meet (authors of papers you cited, potential collaborators). Email them a week before the conference: "Hi, I enjoyed your work on X. I'll be at ICCV and would love to briefly connect. I'll be at my poster on Tuesday or can grab a coffee Wednesday." Specificity gets replies.
- Registration Type: Choose wisely. Early-bird rates save significant money.
| Registration Type | Best For | Key Consideration |
|---|---|---|
| Full Conference (Early Bird) | Most attendees, provides access to all main sessions, workshops, and proceedings. | Mandatory if you have a paper to present. Book as soon as it opens. |
| Student Registration | PhD or Masters students. Requires valid student ID. | Significant discount. Sometimes excludes some social events. |
| Workshop-Only | Those only interested in specific workshops, not the main conference. | Much cheaper, but you miss the plenary talks and main poster sessions. |
| Virtual Access | Those unable to travel. Budget constraint. | You miss all in-person networking, which is 70% of the value for many. |
A Day in the Life at ICCV
Here’s a realistic snapshot from my last attendance.
Morning (8:30 AM - 12:00 PM): Plenary or parallel oral sessions. The coffee is bad and the lines are long. Get there early for popular talks. I often skip the first 15 minutes of a session to have deeper conversations in the hallway instead of rushing from room to room.
Lunch (12:00 PM - 1:30 PM): This is prime networking time. Do not eat alone. Find a group, join a table. The conference-organized lunch (if provided) is chaotic but useful. Better yet, have a plan to meet someone specific.
Afternoon (1:30 PM - 6:00 PM): More sessions or workshops. Poster sessions are gold. Don't just walk by. Engage. Ask presenters: "What was the most surprising result?" "What's the biggest limitation of your method?" This leads to real talk.
Evening (6:00 PM onwards): Official receptions, unofficial lab dinners, or small group outings. The official reception is loud and hard to talk. I prefer smaller, dinner-based gatherings. If you're a student, find out if your lab or university alumni are organizing something.
The physical space matters. Convention centers are huge. Wear comfortable shoes. Hydrate. The "conference flu" is real—bring vitamins.
The Real Networking Strategy Most People Miss
Everyone knows networking is important, but most do it poorly. They hand out business cards like candy or give a 30-second elevator pitch that sounds robotic.
Effective networking at ICCV is about curiosity, not promotion.
Instead of starting with "I work on...", try asking questions about the other person's work. "I saw your poster on domain adaptation. What do you think is the biggest unsolved problem in that area right now?" This opens a genuine dialogue. Your goal is to build 3-5 meaningful connections, not collect 50 LinkedIn requests you'll never follow up on.
The best connections often happen in the margins: waiting in line for coffee, between sessions, at the poster when the crowd thins. Be approachable. Put your phone away.
A specific tactic: Identify the "social hubs." There's always a particular coffee stand or lounge area where people congregate between sessions. Camp out there. You'll overhear conversations and naturally join in.
Follow-up is everything. Within 24 hours of meeting someone, send a short email. "Great chatting about unsupervised detection today. Here's the link to my lab's repo I mentioned." This solidifies the connection.
Your ICCV Questions, Answered Honestly
My paper got rejected from ICCV. Should I immediately submit it to another conference?
Slow down. First, carefully read the reviews. Were the criticisms about novelty, experiments, or clarity? If it's about missing baselines or unclear writing, you can fix that in a month and aim for another venue. If the core idea was deemed incremental, you need a more substantial rethink. Blindingly resubmitting the same text with a new title is a waste of time and harms your reputation with reviewers who might see it again. Take the feedback, let it sit for a few weeks, then decide if it's a revision or a new project.
Is it worth attending ICCV if I don't have a paper accepted?
It depends on your career stage and goals. For a first-year PhD student, it can be overwhelming and expensive. Watching talks online might be more efficient. For someone mid-career looking to build collaborations or transition into industry, the networking opportunities alone can justify the cost. The workshops are also excellent for deep dives. Consider the workshop-only registration as a lower-cost entry point. The main value is in the people, not the proceedings you can read online later.
How do I approach a well-known professor or researcher at the conference without being awkward?
Timing and context are everything. Don't ambush them after their talk as they're packing up (they're tired and everyone wants a piece of them). Instead, go to their poster if they have one, or attend a workshop they're organizing. Ask a specific, intelligent question about their work that shows you've read it. "In your paper on X, you assumed Y. What would happen if Z were different?" This is better than generic praise. If they're surrounded, wait for a lull or send an email afterward referencing the brief interaction. Most are happy to talk to engaged junior researchers.
What's the biggest mistake first-time ICCV attendees make?
Trying to do everything. They sprint from oral session to oral session, exhausted, not retaining anything and talking to no one. You cannot see all the talks. Your brain will melt. Pick 2-3 key sessions per day max. Spend double the time you think you should in poster halls and in conversations. The schedule is a suggestion, not a mandate. The magic is in the unstructured time between events.
The final word? ICCV is a unique beast. It's intense, expensive, and can be exhausting. But done right, it's an unparalleled catalyst for your research, your network, and your understanding of where the field is headed. Go with a plan, focus on people, and don't forget to occasionally step outside the convention center to see the city you're in. It helps with perspective.
This guide is based on firsthand experience and observations from multiple ICCV cycles. Details regarding specific dates, locations, and registration fees should always be confirmed against the official ICCV conference website.
Leave a comment