Computer Vision Toolbox Model for Mask R-CNN Instance Segmentation is not working.

I'm attempting to train the Mask R-CNN on a custom dataset. Mathworks has provided the toolbox for implementing Mask R-CNN. However, when I train the network with my custom dataset, I receive the following error:

[detector,info] = trainMaskRCNN(trainingData,lgraph,options,FreezeSubNetwork="backbone");

↑

Invalid argument at position 1. Expected input to be of size MxNx5, but it is of size 500x724.

I'm not sure what this mistake means. I need clarification and assistance in determining what went wrong here.

Answer 1

[Prince Kumar]

Hi,

The issue here is with your training data dimensions.

As it is clearly stated in the error "Expected input to be of size MxNx5, but it is of size 500x724". There is a dimension mis-match. Please have a look at you data more carefully.

Please go through the following documentation to have a better understanding:

• https://www.mathworks.com/help/vision/ug/getting-started-with-mask-r-cnn-for-instance-segmentation.html
• https://www.mathworks.com/help/vision/ug/example-InstanceSegmentationUsingMaskRCNNDeepLearningExample.html

Hope this helps!