Atlas record: https://www.aitimeline.dev/en/events/evt-0216/
Benchmark · 2010
The first ILSVRC
The ImageNet dataset became an annual competition: a thousand classes, a million and a half images, one metric for everyone.
Why it matters
Computer vision acquired an annual measure, and it is where the advantage of deep networks became visible two years later.
The format was inherited from the PASCAL VOC challenge but the dataset and the number of classes were both enlarged: from 19,737 images and 20 classes in PASCAL VOC 2010 to 1,461,406 images and 1,000 classes in ILSVRC 2010. The first two years were won by methods with hand-designed features: the best top-5 error was 28.2 percent in 2010 (NEC) and 25.8 in 2011 (XRCE). In 2012 AlexNet gave 15.3 percent (16.4 on the supplied data alone) while the best of the rest had 26.2, a gap nobody could dispute. Without this annual ruler the 2012 result would have remained one paper among many.
Event record
- Event date
- 2010
- Timeline date
- Event date
- Verification
- Sources gathered automatically · September 17, 2026
- Lines
- ID
- evt-0216
The first running was in 2010; sources do not give a month.
Related events
Earlier
- Extends ImageNet 2009
The dataset was turned into an annual competition with a fixed metric.
- Related A contest where everyone is scored the same way 2005links to this record
The 50 percent box overlap rule detection was scored by here was inherited by ILSVRC along with the annual contest format itself.
- Related A hundred and one classes instead of five 2004links to this record
The first set on which recognition was scored over a hundred classes; six years later ILSVRC moved the same arrangement to a thousand.
- Related A blind evaluation for face recognition 1996links to this record
The same arrangement fourteen years before ImageNet: a shared set, a withheld test split and an independent scorer.
Later