This is aimed to be a short primer for anyone who needs to know the difference between the various dataset splits while training Machine Learning models. For this article, I would quote the base definitions from Jason Brownlee’s excellent article on the same topic, it is quite comprehensive, do check it out for more details. Training DatasetTraining Dataset: The sample of data used to fit the mode

Spatial transformer is yet another LEGO block in the collection of differentiable modules. It removes spatial invariance from images by applying a learnable affine transformation followed by interpolation. The STN block can be placed in a convolutional neural network (CNN) and it works mostly by itself. The full code for this project is available on GitHub: https://github.com/dnkirill/stn_idsia_co

Deep neural networks and Deep Learning are powerful and popular algorithms. And a lot of their success lays in the careful design of the neural network architecture. I wanted to revisit the history of neural network design in the last few years and in the context of Deep Learning. For a more in-depth analysis and comparison of all the networks reported here, please see our recent article (and upda

Squeeze-and-Excitation Networks (SENets) introduce a building block for CNNs that improves channel interdependencies at almost no computational cost. They were used at this years ImageNet competition and helped to improve the result from last year by 25%. Besides this huge performance boost, they can be easily added to existing architectures. The main idea is this: Let’s add parameters to each cha

Example of end-to-end object detection (from Microsoft)This post is meant to constitute an intuitive explanation of the SSD MultiBox object detection technique. I have tried to minimise the maths and instead slowly guide you through the tenets of this architecture, which includes explaining what the MultiBox algorithm does. After reading this post, I hope you will have a better grasp of SSD and wi