How do I calculate the precision, recall and F1score for an image database?

I work on dermoscopy images, I used fcm clustering for segmentation and also I obtained another mask to overlay on the initial image and I need to see if the segmentation was good enough. Can you please provide some ideas? Thanks in advance

Answer 1

[KALYAN ACHARJYA]

To calcuate the precision and recall you need ground truth images

Lets clear the issue

1. You have dermoscopy images
2. You applied some method to finding region of Interest>> as result segmneted image
3. Now you have to compare resultant segmented image with respective Grond thruth image (or Manual verification by experts)

Precision=TP/(TP+FP); I assuemd that you know these terms

Recal=TP/(TP+FN)

F1 Score= Average of 1/(Precision & Recal)

Must read article here (Easy and Best Way)

Regrads

[Neon Argentus]

F1 measure should be average of inverses of precision and recall.

[KALYAN ACHARJYA]

Yes, thanks

[benarabi tarek]