ray's profile - activity

I have successfully implemented TLD in openCV. Is there anyway I can train the detector once, save what it learned on to a file and then use the file to initialise the tracker at some other point of time for further tracking?

I am trying to implement a TLD real time tracker on Open CV. However I couldn't find any concise tutorials as to how to correctly implement a tracker. Please guide me as to how to do so.

Is there anyway to store the learned data to a data file or something like that so that next time the tracker is done, it reads the data from the configuration file and hence it is not necessary to draw bounding boxes around the object of interest every time?

I also wish to be able to detect multiple objects using TLD. Say the detector was first trained to detect cats and then cars. Assume that the detector saved the learned data to files cats.data and cars.data respectively. Is there anyway to initialise the tracker with both these config files so that it detects both the objects?

I am working on an application that is required to detect statues, paintings etc. Which method is the best to detect these objects? I tried this method. However this technique is not accurate . I also tried the Haar Cascade classifier. However the application that I am working on is large and hence it is required to train very large number of such images, which will be very tedious if we proceed with Haar Cascades.

Is there any other method that can detect objects with reasonable accuracy and at the same time quick enough? Even if so how can one detect multiple objects( multiple paintings, statues etc) ?