Ask HN: Why aren't convolutional decision trees or convolutional SVMs a thing?

Put another way, what made ANNs more suited to convolutional feature extraction from imagery (vs. other ML methods)?