Using Image Labeler to train Object Detector

Asa Kalish

2018 6 17

0 个回答

2 次查看（30 天）

0 个投票

test.png

 dataDir = fullfile('/Users/asakalish/Documents/MATLAB/Image Analysis/');
imDir = fullfile(dataDir,'data');
pxDir = fullfile(dataDir,'dataPixelLabels');
imds = imageDatastore(imDir);
classNames = ["1" "2" "3" "4" "5" "6" "7" "8"];
pixelLabelID = [1 2 3 4 5 6 7 8];
pxds = pixelLabelDatastore(pxDir,classNames,pixelLabelID);

This is the code I have so far

Attached is an example of the images I am training the algorithm to detect. I want the algorithm to be able to find each white "dot" (they are organelles), and effectively draw a circle around them so that I can analyze them and take data such as centroid, size, etc.

I tried to upload them into Image Labeler, to identify each organelle, and put them into the 'dataPixelLabels' folder. However, after doing the ROI's for each image, I don't know how to export them into the 'dataPixelLabels' folder, as a .png/.jpg/etc

I think I need to label these images and put them in this file, to be able to train the algorithm. I have not worked with CNN's before, so this is new to me. If you believe this is the wrong approach, please let me know. Any help appreciated!