I want to Implement a Image classifier which is able to detect defects from a particular set of smaples and defects are unknown.
We have both good and bad samples, but there are no features defined. The Images are of 400*400 size.
The images are taken as a tiff file containing 24 files of a single sample from 24 different random positions.
Which machine learning techniques could be applied?
Thank you
It depend on many factors such as amount of datasets that you have. If you have much enough dataset you can choose CNN, otherwise you might use traditional machine learning such as SVM.
High resolution it may not impact on classifaction, but SVM give satisfactory result
Shounak Datta
Thank you for your response, but There are NO labels defined.
Khairun Saddami Thank you for your response but could you suggest me some resources which could help me approach the problem practially(coding examples) and Theoretically to build a proof of concept.
You can search it from github
or in mathwork.com
just search what do you want, CNN or SVM implementation
Dear Manav,
Since you have no labels available, you have to use clustering techniques with autoencoders to learn the feature representations. Have a look at https://arxiv.org/pdf/1610.04794.pdf
- Badges
- Science topic
- Similar topics
- Computer Science
- Images