Skip to content

3 Reasons to Visit Voxel51 at ICCV23!

As Jacob Marks said in his recent post, ICCV 2023 Survival Guide: 10 Computer Vision Papers You Won’t Want to Miss, ICCV is shaping up to be something special! 

The premier international computer vision event is happening this week in Paris, France, with the main conference and expo taking place on October 4-6. Events like this are an amazing way to discover new ideas and build connections with others in the computer vision community.

Here are three reasons to put a visit to Voxel51 Booth #55 on your ICCV23 agenda!

1 – 💡 See visual data in a whole new light, including six new ICCV datasets!

Nothing hinders the success of machine learning systems more than poor quality data. But how do you see issues hiding in your datasets that are limiting your ML model performance? Shine a light into your visual data and your model’s failure modes using open source FiftyOne!

FiftyOne provides unprecedented visibility for optimizing your dataset analysis pipeline. Use it to illuminate visual data in ways that matter most to you: visualize complex labels, explore samples and scenarios of interest, find annotation mistakes, evaluate your models, identify model failure modes, and much more!

While it’s possible to load any visual dataset into FiftyOne, we loaded up these ICCV datasets so you can instantly see for yourself. Visit Voxel51 at ICCV Booth #55 to see these datasets in action, or explore them now in your browser!


Building3D is an urban-scale dataset and benchmarks for learning roof structures from point clouds from this paper by Ruisheng Wang, Shangfeng Huang, and Hongxin Yang. Building3D consists of more than 160 thousand buildings along with corresponding point clouds, mesh and wire-frame models, covering 16 cities in Estonia – about 998 Km2.

Explore Building3D instantly in your browser! >>


EgoObjects is a large-scale egocentric dataset for fine-grained object understanding published in this paper by Chenchen Zhu, Fanyi Xiao, Andres Alvarado, Yasmine Babaei, Jiabo Hu, Hichem El-Mohri, Sean Chang Culatana, Roshan Sumbaly, and Zhicheng Yan. EgoObjects 1.0 includes 114K annotated frames (79K train, 5.7K val, 29.5K test) sampled from 9K+ videos collected by 250 participants from 50+ countries using 4 wearable devices, and over 650K object annotations from 368 object categories.

Explore EgoObjects instantly in your browser! >>


EqBen (Equivariant Benchmark) is a new challenging benchmark created to diagnose the equivariance of VLMs with visual-minimal change samples. It was published in the paper Equivariant Similarity for Vision-Language Foundation Models by Tan Wang, Kevin Lin, Linjie Li, Chung-Ching Lin, Zhengyuan Yang, Hanwang Zhang, Zicheng Liu, and Lijuan Wang.

Explore EqBen instantly in your browser! >>


Mose is a new dataset for video object segmentation in complex scenes published in this paper by Henghui Ding, Chang Liu, Shuting He, Xudong Jiang, Philip H.S. Torr, and Song Bai. MOSE contains 2,149 video clips and 5,200 objects from 36 categories, with 431,725 high-quality object segmentation masks. The most notable feature of the MOSE dataset is complex scenes with crowded and occluded objects.

Explore MOSE instantly in your browser! >>


SatlasPretrain is a large-scale dataset for remote sensing image understanding, published in this paper by Favyen Bastani, Piper Wolters, Ritwik Gupta, Joe Ferdinando, and Aniruddha Kembhavi. SatlasPretrain combines more than 30 TB of satellite images from public sources such as Sentinel-2 and NAIP with 137 label categories, making it an effective pre-training dataset that greatly reduces the effort needed to develop robust models for downstream satellite image applications.

Explore SatlasPretrain instantly in your browser! >> 


SportsMOT is a large multi-object tracking dataset in multiple sports scenes, published in this paper by Yutao Cui, Chenkai Zeng, Xiaoyu Zhao, Yichun Yang, Gangshan Wu, and Limin Wang. SportsMOT consists of 240 video sequences, 150K+ frames, and 1.6M+ bounding boxes from basketball, volleyball, and football.

Explore SportsMOT instantly in your browser! >> 

2 – 👋 Meet fellow enthusiasts in the computer vision community

We love meeting fellow members of the computer vision community! Come meet ML engineers, developers, co-founders, and open source enthusiasts from the Voxel51 team at booth #55. Let’s discuss all things computer vision, explore FiftyOne hands-on, and send you home with our latest epic swag (while supplies last!).

3 – 👕 Score epic FiftyOne swag

Voxel51 is known for bringing epic swag to community gatherings. And, who doesn’t like a good giveaway to commemorate an amazing event? Stop by our booth to see our lineup of goodies, including a Paris- and ICCV23-inspired limited-time tshirt! And snag some swag for yourself! 

As much as we love swag, we also simply love welcoming new members to the open source FiftyOne community. Why? We know how valuable FiftyOne truly is for data engineers and scientists and therefore getting it into the hands of even more people is what we’re all about. 

Even if you come for the swag, we’re convinced you’ll walk away with two things you’ll love: sweet gear and first-hand experience of how FiftyOne can help you level up your computer vision workflows.

Can’t make ICCV? 

If you can’t attend ICCV this year, no problem! Here are other ways to connect with the FiftyOne community: