search

YunchaoYang / Blogs

blogs and notes, https://yunchaoyang.github.io/blogs/
0 stars 0 forks source link

Binary classification vs multi-class classification vs Multi-label Classification #38

Open YunchaoYang opened 1 year ago

YunchaoYang commented 1 year ago
Binary classification multi-class classification Multi-label Classification Multi-output Classification
Sigmoid classifier and BinaryCrossEntropy Loss Softmax classifier and CrossEntropy Loss Sigmoid classifier and BinaryCrossEntropy Loss Sigmoid classifier and BinaryCrossEntropy Loss
----------------------- ---------------------------- ----------------------------

Note the difference between the Multi-label and Multi-output classification: For Multi-output, each entry of the output has a fixed meaning of 0/1. Each class must choose from 0 and 1. And the number of output is fixed.

For multi-label, you have a pool of output labels, you are asked to choose the proper labels from the label pool that described in the image. The label pool is usually large and abundant.

reference:

https://learnopencv.com/multi-label-image-classification-with-pytorch-image-tagging/