Object tracking is pretty easy conceptually. And if you have a good detector, simple methods can be pretty effective.
Cross-entropy loss in PyTorch
Cross-entropy loss in PyTorch can be a little confusing. Here is a simple explanation of how it works for people who get stuck.
Pairwise distance in NumPy
You can do vectorized pairwise distance calculations in NumPy (without using SciPy). This lets you extend pairwise computations to other kinds of functions.
Everything is a pipeline
DAGs are everywhere in scientific computing and understanding them is key to running large computations efficiently.
Use informative names
Informative names make programming easier. Here are a few suggestions you can use to write better names in your code.
Download a YouTube video from the command line with youtube-dl
You can use youtube-dl to download arbitrary YouTube videos from the command line. This is useful for doing machine learning on new videos.
Machine learning pipelines should be Python packages: Transfer learning
Building machine learning pipelines as well-formed Python packages simplifies transfer learning. Here's a simple example.