The goal of this Kaggle competition is to remove the background of a set of car pictures with a width variety of year, color and model combinations. That means, creating a mask for each photo that covers the area where the vehicle is. Using Machine Learning in this task would save a lot of time in manual photo editing.

From speech recognition to music generation, a lot of work has been done in the field of Machine Learning applied to Signal Processing. The project I present here is a bit simpler than that but still very useful. Synchronizing audio and subtitles is a tedious and boring task since you have to identify when the human speech happens and modify each subtitle to fill that temporal space. But, what if a machine could do that for us?

To deal with it, in this post I describe how I have trained a Neural Network to detect human speech on videos and use that information to synchronize their subtitles.

Finding repeated questions in discussion forums is fairly common and that can be a problem. Multiple similar questions lead seekers to find the best answer in different posts, and writers to answer the same question multiple times. So, to improve the user experience it is necessary to avoid these repeated questions through paraphrase identification.