Limit on images in object recognition module
Jeremy from United States  [12 posts]
2 years
I'm attempting to use the object recognition module for face tracking, training faces in different poses. This will take quite a few trained images, I think, and I understand there is a limit to how many images can be learned by one module. How many objects can be trained, and how do you tell when you have reached the limit? Does the module no longer find any matches? Thanks in advance!
Steven Gentner from United States  [260 posts] 2 years
The limit is 1000 images on the Object Recognition module.

But you'll probably never get that many loaded. The issue is that thing WILL continue to work, just very VERY slowly. (Yes, even on an I7) The more images you load, the slower the recognition process will be. You will probably also run out of memory depending on how large the template files are since everything needs to fit into memory too.

So it WILL work, just VERY slowly.

Typically you instead use a generalized model like the Face detection module to eliminate most of the image and run a specialized recognition over the results of that module.

Are you intending to use the templates to identify a specific person or just people in general?

Anonymous 2 years
I would like to be able to recognize specific persons, if possible, but I would settle for identifying people in general. The trouble I have run into so far is, I am not able to track a face in profile. Some of the software solutions I have tried thus far, including RR, do pretty good at identifying a frontal face, but they can't track a face from the side.
Is it still a possibility that you could build a module that could be trained to recognize faces in profile? If so, I would be interested in hearing more about it. Thanks.

This forum thread has been closed due to inactivity (more than 4 months) or number of replies (more than 50 messages). Please start a New Post and enter a new forum thread with the appropriate title.

 New Post   Forum Index