Computer Vision News - January 2019

The tip of this month will review two tools to help you train your neural network more systematically. These tools help you follow all the training processes and results in one place in a very user-friendly way. The first tool is actually a website -- comet.ml -- which interfaces with your code to enable you to see the outcomes of your runs online. The second tool is called Hyperas and it allows you to define what network configurations you would like to run and test, using standard deep learning libraries, such as Keras. Let’s start with comet.ml Comet enables you to track your Machine Learning experiments, facilitating comparisons you may want to make and collaborations. It allows tracking hyperparameters, metrics, code, stdout, etc. and supports standard libraries -- Keras, TensorFlow, PyTorch, scikit-learn out of the box, and other libraries with the manual API. Let’s try it out to see how it works. First, you need to install the package locally, on your own computer, like this: To use comet.ml, you need an api-key , which you’ll receive after you set up an account on their website. You use your api-key as an Experiment class parameter -- as seen in the snippet below. This is what links your code to the website. 20 Tip - Train Your Network Focus on by Assaf Spanier These tools help you follow all the training processes and results Computer Vision News pip install comet_ml from comet_ml import Experiment experiment = Experiment(api_key="YOUR_API_KEY") # Your code.