- The output of face detector is not always the same, it can be a square, a rectangle, or an oval bounding box.
- Most of the landmark detectors need to take in an square bounding box for the detection.
- Although the bounding box shape is different, they roughly have the same shape center. For the square and rectangle, they have the same bounding box center, and the edge length of the square box is roughly the same as the mean value of the two edge lengths of the rectangle. Here is a sample.
The repository cloned from GitHub pytorch/pytorch is different from the package we download using
pip install or
conda install. In fact, the former contains many C/C++ based files, which consist of the basic of Pytorch, while the latter is more concise and contains compiled libraries and dll files instead.
Here, let’s discuss the release version, or the installed package at first. The package has a lot of components, Here I only pick out some most important parts to do explanation.
Video classification, or in our case, more specifically, action recognition, are studied for a long time. There are many traditional as well as deep learning based method developed to address this problem, and the latest action recognition result trained on a large dataset Kinetics can even reach 98% accuracy. Considering the fact that the action we need to classify is not too much, giving enough data and using the pre-trained model on Kinetics, the result can be quite promising.