Creating a mask detector and whatsapp alarm system with a raspberry pi camera
We all should be wearing masks. What about training a model to identify if you are wearing yours and, in case you are not, send you a message to remind you t...
Recent Posts
Building a prediction API using FastAPI and Docker
Most of the times your model is used behind the scenes, interacting with other components and providing its results through APIs. Let’s see how to do it!
Using convolutional neural network’s feature vectors to find similar images
If you have a well trained model, it should produce a robust vector representation of your data and similar observations should be next to each other.
Explaining predictions using shap
A lot of times in consulting clients want to understand why a model is giving a prediction. Shap is a fantastic library to help dealing with that.
Training multi-task convolutional neural networks to save GPU costs
Having many convolutinal neural networks can get expensive due to GPU costs. Let’s save money and time by making them multi-taks!
Retro style image quantization with k-Means
Another educational post - I needed to explain k-means to some students and wanted a different use case to make it more engaging.
Getting spotify tracks features and classifying them with tensorflow
I needed to explain the basics of neural nets to some students. Thought it would be interesting and engaging to use spotify track features to do it!
Vectorizing images using pre-trained convolutional neural networks
Today we can easily leverage the most powerful convolutional neural networks to vectorize our images and use them for any application we want! Exciting!