If you want to create a machine learning model but say you don’t have a computer that can take the workload, Google Colab is the platform for you. Even if you have a GPU or a good computer creating a local environment with anaconda and installing packages and resolving installation issues are a hassle. Colaboratory is a free Jupyter notebook environment provided by Google where you can use free GPUs and TPUs which can solve all these issues.
After we trained our first machine learning using basic foundation, it’s time to use slightly advanced tool to tackle more challenging problems. The concept is technically the same; we have to feed the data, normalize it, and send the knowledge through the network back and forth until the loss function is minimized.
Unsupervised learning is a type of self-organized learning that helps find previously unknown patterns in data set without pre-existing labels. It is also known as self-organization and allows modeling probability densities of given inputs. It is one of the main three categories of machine learning, along with supervised and reinforcement learning. autoencoder is the main component that is used in unsupervised learning.
Convolutional Neural Networks are very similar to ordinary Neural Networks from the previous chapter: they are made up of neurons that have learnable weights and biases. Each neuron receives some inputs, performs a dot product and optionally follows it with a non-linearity. The whole network still expresses a single differentiable score function: from the raw image pixels on one end to class scores at the other. And they still have a loss function (e.g. SVM/Softmax) on the last (fully-connected) layer and all the tips/tricks we developed for learning regular Neural Networks still apply.
In this series of lectures, we covered the basis of machine learning, methods and designs, problems and many solutions to overcome training difficulties. In this session we will go a little further and discuss the master models that scientists have developed and how to use them to solve many of the challenges in the wild. Let’s get started with this new topic using this Colab.
Deep learning algorithms have solved several computer vision tasks with an increasing level of difficulty. We previously tackled image classification and image reconstruction and today we tackle object detection. The image semantic segmentation challenge consists in classifying each pixel of an image (or just several ones) into an instance, each instance (or category) corresponding to an object or a part of the image (road, sky). This task is part of the concept of scene understanding: how a deep learning model can better learn the global context of a visual content. Let’s get started with this new topic using this Colab.