I have to build a madel which could classify an image as good or bad at the end.
what should be the ideal resolution and pixel size for the task.
Do we need to apply image segmentation techniques to just take the required region from a whole image. or the whole image could be fed into the neural network.
I have a particular object that I need to classify as broken or not broken inside a whole image.
It depends !!! I would take the ROI and then feed it to CNN else CNN will take you for a toss with feature maps and filters ...
Dear Mr. Madan,
The higher the resolution the better. Of course you need to do segmentation to separate objects and the better the segmentation results the better is the recognition results. You can start with Self-Organizing Maps as an unsupervised ANN method to segment the image and then use another algorithm to classify based on a reference map or object. The goal of the segmentation process is to change the characteristics of the image into more meaningful ones, thus facilitating interpretation and classification. Because these image segments better represent objects in the landscape than do the original pixels, each step of the classification process, from defining training sites to classifying from these segments, is simplified. It is also possible to achieve better accuracy. The common salt-and-pepper effect that results from a pixel-based classification is reduced and a more cartographic-grade map is the result.
Thank you so much, for your valuable suggestions.
M. Awad and Aparna Sathya Murthy
- Badges
- Science topic