liveBook · Manning

Mean pixel IoU

As explained in Sutskever's List

Mean pixel IoU is the average intersection-over-union between predicted and true regions across pixels or semantic classes in an image-segmentation evaluation.

Mean pixel IoU summarizes how much predicted object regions overlap ground-truth regions pixel by pixel. It is a common benchmark for semantic segmentation. section 3.5

Comparing segmentation models

On PASCAL VOC, a dilated-convolution model outperformed leading models by more than five percentage points in mean pixel IoU. section 3.5

How Sutskever's List covers it

How deep it goes
Defines pixelwise overlap but gives no calculation procedure.
Examples
Compares segmentation models on PASCAL VOC and reports a gain exceeding five percentage points.
What you can do after
Evaluate segmentation performance by comparing predicted regions with ground-truth regions.
How it connects
Connects semantic segmentation and dense prediction through pixel-level region overlap.
Cover of Sutskever's List
liveBookmanning.com
Read Sutskever's List on liveBook Get the book at manning.com

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.