Battle of Learning Machines - Watson vs. Tensorflow on Image Classification
Out of so many machine learning algorithms that exist today, which do you think is easier to work with? Which one is better at visual recognition and more specifically, at image classification?
Today I test two of them: IBM's Watson and Tensorflow (with its powerful support from Google). The task is to classify three images: an image of a piece of broccoli, an image of a castle, and an image of a swan. You'll have to watch the video below to know my verdict.
A few things to point out:
- Watson works by making API calls (so it needs an internet connection). There's a limited number of free calls/day.
- Tensorflow is powered by a pre-trained classifier (a deep convolutional neural network - Inception - that has been trained on the ImageNet 2012 Challenge dataset and it can categorize images into about 1000 classes or categories)
- Tensorflow works locally