Skip to content

Tips And Tricks

Tim Büchner edited this page Apr 18, 2024 · 3 revisions

Bounding Box

As we have discussed in the Model and Methods wiki entry, we have two face detection algorithms running in the background. Accordingly, the manual default bounding box can definitely be bigger than the actual face, as the second face detector focuses on the face. This is especially useful if you know your person will move to a certain area. Then, you can adapt the bounding box to cover this whole area. If you expect the whole frame to be used, you can disable the usage of the bounding box in the GUI menu and only the second face detector is activated.

Video Conversion

First of all, we recommend you use the original obtained video and refrain from using any conventions in different file formats. As compression algorithms like H264 or H265 often replace subsections of the image from previous frames, we observed that in videos with high temporal resolutions as the face is affected. However, the results do not drastically affect any of the extraction capabilities; they could reduce the fine-grained information of the intricate blinking behavior.