image processing - OpenCV: Improving the speed of Cascades detection -
I have to find people to use OpenCV Cascades in real time. Currently I am using trained cascade files which come with OpenCV but later I would train my own LBP cascades to gain mobility. I have a question.
What are the ways to speed up Cascade Detection? For example, see it is really fast, Higher Cascade and makes good use. What types of things can I do to achieve this speed, especially for a real-time application? Any trick and hacks?
I'm not sure what to expect from "speed" in your video example It is difficult to understand how "speed" has been done there. In computer terms, when we talk about the "speed" of detections, we generally mean the frames per second (FPS) or millisecond run-time of the algorithm for each other or video set. If the FPS input received by the algorithm is similar to the FPS of the video, it is called real-time or 1x processing speed. If the process of FPS is more than input FPS, then you are faster than real-time processing and if it is small, then you are slower than real time. Assume when you said "speed" you meant. Given this, I give you two ways to speed detections. I really suggest reading these two papers that have actually been introduced in the past several years in pedestrian identification: and, in both traditional identification settings, many halves calculate the detection of scales. The latter is the publicly available code and But it is therefore one of the areas to improve: scale size . The method originally implemented in OpenCV is based on the Voila-Jones method, which is a high-like feature set used in the search. Another area of improvement is called Windowing . OpenCVV requires basic methods of traditional detection, including one implemented originally so that you slide the scale scale across the image, generally the upper- To go around the left-to-right line-by-line, it is called a classic method (ESS) that optimizes many extensions from this, but to find out the object This is a great place to start and understand things. Now, of course, a very obvious way to speed up the detection process is parallel. Your code, eg multi-threading or GPU, has many GPU implementations that are publicly available, e.g. Using a support vector machine-based detector.
Comments
Post a Comment