In pattern recognition problems,
when we use the sliding window technique to extract features from an image,
Is there an impact of the size of the sliding windows (width and overlap) on the results of the recognition? and why?
thank you in advance.
This is an experimental result that can vary according to the problem.
You can compare the results by trying many sliding windows.
Because this process is problem-dependent. Depending on the problem, the effect of the sliding window width and overlap will change.
The sliding window method helps hold context in the sensor values within the window and the amount of overlap keeps the data windows interdependent or have some similarity/continuity in terms of feature. Both window size and overlap has effect on the classifiers accuracy. It is quite common in classifier design to have windows created to extract feature with 50% overlap.
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts