It still is super relevant. Most computer vision done outside academia is still based on older stuff, or classical computer vision algorithms. You don't really get so many chances to use the latest models and techniques, as most often than not, they are not that relevant, or are only for extremely specific cases, or you just don't need something that complex.