Exploring the scope of Automation and Artificial Intelligence in Image data
I discuss about information, which can not be found in a picture, especially the contextual information. Later, I have mentioned some business benefits from taking such an approach.
The purpose of this article is to give direction of what next can be done in the field of image analysis and how to consider context, and to answer such question as: “Can we consider human inputs to marry Automation and AI?”
Looking at the picture above (figure 1), could anybody have predicted what is the essence of the story? The context is that the skiing partner of Sebastien de Sainte Marie had been taken away by an avalanche a few meters away from a couloir that they both had skied together the day before, which had prompted Sebastien to ski alone. Perhaps, the prediction was not so difficult because the only thing being shown in this picture is a man staring several feet (and maybe hundreds of feet) down the mountain gorge. …
Neural networks were arguably losing momentum and were almost getting forgotten and were sinking like a Titanic ship, until Support Vector Machines came to their rescue. Not long ago, SVM started to flounder on some problems related to solving non-linear relations in data.
SVM outperformed NN in protein-fold recognition, and in time-series forecasting. SVM outperforms NN as the size of the training data decreases. Consider that multi-layer perceptrons (MLP), though they are often excellent classifiers, are driven by a numerical optimization routine, which in practice rarely finds the global minimum; moreover, that solution has no conceptual significance. On the other hand, the numerical optimization at the heart of building an SVM classifier does in fact find the global minimum. …