#Machine Learning Assignment: Solving Synthetic Visual Reasoning Test(SVRT) with Few-shot Learning Algorithms
This series of synthetic image recognition problems were developed to assess the performance of machine-learning techniques for vision. They are designed to require more than local descriptors and simple statistics of the image to be solved properly.[http://www.idiap.ch/~fleuret/svrt/]
Previous studies show that the 'same-different' problems from SVRT are chanllanging to solve through supervised learning, on the contrary, humans are easy to detect this kind of abstract feature by observing only a few samples.
(Same-Different)Problems 1, 5, 6, 7, 8, 15, 17, 19, 20, 21, and 22 involve comparing shapes. An agent has to be able to decide whether two shapes are similar or not at one stage of the classification process.