ImageNet Large Scale Visual Recognition Challenge

As explained in Sutskever's List

The ImageNet Large Scale Visual Recognition Challenge is an image-classification benchmark and competition based on the ImageNet dataset.

The ImageNet Large Scale Visual Recognition Challenge (ILSVRC) was an image-classification competition where models were compared by top-5 error: how often the correct answer was absent from a model’s five highest-ranked guesses. section 2.0

In 2012, AlexNet achieved 15% top-5 error, compared with about 26% for conventional methods. CNNs then dominated the competition: by 2014, the winning error rate was 6.7%, and by 2017, 29 of 38 teams exceeded 95% top-5 accuracy. section 2.5.5 section 2.0

Figure 2.6 The top-5 error rates in the ILSVRC from 2010 to 2017. SIFT and SVM won in 2010 and 2011, but performance plateaued. CNNs won from 2012 to 2017, including AlexNet in 2012, and eventually saturated the benchmark.

Compare top-5 error rates across the competition’s 2010–2017 results, from early SIFT and SVM entries to CNN winners. section 2.5.5

How Sutskever's List covers it

How deep it goes

Traces competition results from 2010 to 2017 but does not detail evaluation procedures beyond top-5 error.

Examples

Compares AlexNet’s 15% top-5 error in 2012 with conventional methods’ roughly 26% and GoogLeNet’s 6.7% in 2014.

What you can do after

Evaluate how top-5 error comparisons tracked the shift from handcrafted features to CNNs.

How it connects

Connects ImageNet and AlexNet with convolutional neural networks, SIFT, SVM, and PASCAL VOC.

liveBookmanning.com

Read Sutskever's List on liveBook Get the book at manning.com

Related concepts

• AlexNetaddressed by · AlexNet is a convolutional neural network architecture designed for image classification and recognized for its landmark performance in large-scale image recognition.
• labelproduces · A label is a value or annotation associated with an item that identifies its target category, outcome, or reference answer for learning or evaluation.
• ResNetmeasured by · ResNet is a family of deep neural-network architectures built from residual blocks with shortcut connections that facilitate information flow and optimization.

This explainer was generated by Manning's knowledge graph from the text of Sutskever's List. The section links point to the parts of the book it draws on.