: With a formidable 2359 papers accepted out of a whopping 9155 submissions, the conference is about to be a marathon of mental stimulation. This volume speaks volumes concerning the explosion of ideas in the sphere.

: The typical CVPR paper has roughly 5.4 authors, highlighting the collaborative nature of the sphere. The ability of teamwork is on full display within the paper with probably the most authors — “Why is the winner the very best?”, boasting a whopping 125 contributors. However, the conference also celebrates individual brilliance with 13 papers penned by a single writer.

: Among the many accepted papers, 1724 have versions on Arxiv, and 68 additional papers were found elsewhere, showing that the research community extends beyond traditional academic boundaries.

: The words “dataset” and “model” were a dynamic duo, appearing together in 567 abstracts. “Dataset” featured solo in 265 abstracts, while “model” took the lone highlight 613 times. Clearly, the interplay between data and modeling is a point of interest in computer vision research.

: In line with CVPR paper abstracts, the most well-liked datasets this 12 months were ImageNet (105), COCO (94), KITTI (55), and CIFAR (36). These datasets are just like the sandboxes where researchers play to create latest models and algorithms.

: Of the 2359 papers, 1487 titles had acronyms or compound words with multiple capital letters. That’s a staggering 63%! From CLAMP and PATS to SIEDOB and FJMP, the acronyms added a splash of fun (and sometimes confusion) to the combination.

That is CVPR 2023 in numbers, an event that’s as much about data because it is about computer vision. These stats give us a snapshot of the various, collaborative, and vibrant research landscape of this field. Able to dive deeper? Let’s explore the new trends and key themes of this 12 months’s conference next.