OpenCV is the computer-vision library that has been the standard since 2000, with bindings for C++, Python, Java, and (via OpenCV.js) the browser. Most courses use the Python interface because it's the easiest entry point, but production systems often switch to C++ for performance.
The library covers classical computer vision (image manipulation, feature detection, tracking, calibration, stereo vision) plus a thin layer over modern ML models (DNN module for running pre-trained ONNX or TensorFlow models). For pure deep-learning vision work, PyTorch or TensorFlow are usually the better choice; OpenCV's sweet spot remains the preprocessing and post-processing layers around ML models.