1

Working on an image segmenetation problem, I've tackled the following scenario repeated on different images:

High Recall and Accuracy (around 99%) Low IoU (around 60%)

How is that possible?

Recall is basically the rate of recognizing the specified class (let's consider a single class problem) and Accuracy is the rate of missing the specified class (well, the opposite).

IoU basically takes both measures into account, how come this IoU rate is that low compared to the above?

Thank you.

Jes
  • 11
  • 1
  • I am not into image-segmentation but I can tell you that accuracy is the proportion of all right predictions to the number of predictions. High recall AND accuracy is therefore very well possible. – Pedrinho Aug 19 '19 at 17:51
  • What is IoU? How these are possible depends heavily on the base rate. See [Why is accuracy not the best measure for assessing classification models?](https://stats.stackexchange.com/q/312780/1352) The criticisms listed there apply also to recall (and I suspect to IoU). – Stephan Kolassa Aug 19 '19 at 17:53
  • I thought you should first know the four fundamental concepts: recall, precision, F measure and Jaccard index. – Lerner Zhang Aug 19 '19 at 22:57

0 Answers0