Your submission is now in Draft mode.

Once it's ready, please submit your draft for review by our team of Community Moderators. Thank you!

Submit Essay

Once you submit your essay, you can no longer edit it.


This content now needs to be approved by community moderators.


This essay was submitted and is waiting for review.

Image Classification Index 2022-01-14


This question is part of the Hill Climbing Round of the Forecasting AI Progress Tournament. You can view all other questions in this round here.

Image classification is the task of identifying an image by assigning to it a specific label. Typically, Image Classification refers to images in which only one object appears and is analysed. In contrast, object detection involves both classification and localisation tasks, and is used to analyse more realistic cases in which multiple objects may exist in an image.


The index is constructed as follows:

  • We take the average (arithmetic mean) of - ln (error) of the state-of-the-art performance across all benchmarks in the index
  • The index is then defined by scaling this mean so that its average value for the year 2019 is 100

The following benchmarks are included in the Image Classification Performance Index:

Image classification on: ImageNet (in top-1 accuracy), STL-10, CIFAR-100, SVHN, MiniImagenet 5-way (1-shot), Tiered ImageNet 5-way (1-shot), CUB 200 5-way 1-shot, Stanford Cars, CUB200, FGVC Aircraft

Historical data on the Image Classification Performance Index may be found here. As of writing this question, the index is at 114.88 for December 2020.

What will the value of the herein defined Image Classification Performance Index be on 2022-01-14?

This question resolves as the value of this index on 2022-01-14, 11:59PM GMT.

Models that are trained on multiple datasets do not qualify for the purpose of this question—only models trained on benchmark-specific datasets will be considered.

A benchmark will be removed from the index if:

  • At the time of resolution no new performance data is available for new models for the specific benchmark over the previous 6 months
  • The value of - ln (error) for that benchmark exceeds 10

If a benchmark is removed from the index, the index shall simply be re-constructed according the procedure outlined above.

Performance figures may be taken from e-prints, conference papers, peer-reviewed articles, and blog articles by reputable AI labs (including the associated code repositories). Published performance figures must be available before 2022-01-14, 11:59PM GMT to qualify.

In case error is not natively reported, it is constructed by taking 1-accuracy/100, or 1-(percentage correct)/100.

For the purpose of this question, the SOTA models in 2019 represent in the linked Google sheet are assumed to represent the ground-truth, and to maintain consistency, these won't be revised in case these are found to be erroneous or invalid.

Make a Prediction


Note: this question resolved before its original close time. All of your predictions came after the resolution, so you did not gain (or lose) any points for it.

Note: this question resolved before its original close time. You earned points up until the question resolution, but not afterwards.

Current points depend on your prediction, the community's prediction, and the result. Your total earned points are averaged over the lifetime of the question, so predict early to get as many points as possible! See the FAQ.