Someone else was asking about face recognition databases this weekend, so I will point you to that discussion in case there is something in it that helps you.
Thanks MartyG.Thanks for providing the link.
I am sorry it doesn't work for me.
I've tried a lot of ways but they can't solve the problem.
I suspect that you'd have better luck using '2016 R2', since 2016 R3 is still relatively new and it has a very different structure from 2016 R2 and earlier SDKs. So there are far more unknowns about its operation than there are with R2, and it has a lot less in common than the SDKs that were originally used to create documentation for face recognition back in 2015 / 2015.
Getting face recognition to work is at least not as difficult a it used to be, thanks to the body of knowledge that has built up about it over time. For example:
- We now know that errors related to instructions such as CreateStorage (which are in the documentation but not actually implemented in the SDK) can be fixed by 'commenting out' those lines in the scripting.
- We also now know about the existence of the 'On_Demand' face registration mode. Previously, it must have seemed that 'Continuous' was the only type of face registration mode available.
I am sorry,your suggestion has no help.
The problem hasn't been solved for a long time.....
Thanks for taking time to reply........
I researched your problem further, as I want to help you find a solution.
An Intel employee who was posting for help with their project was using the FaceData RecognitionModule.DatabaseBuffer instruction to save face data. They got saved data to load back in the next time the project was run by saving the face data to a .dat format file. By doing this, the face data automatically loaded into the project the next time it was started. I do not know how you would save in .dat format though - perhaps someone else on this forum can offer advice on that.
In another post, Intel employee Gal G suggested to someone with a null face reference that they should only do configuration once per frame, and before acquiring the frame.
Thank you very much for your help.Can you help me ask the specific code or how do they configure FaceData.
They didn't provide code along with their message, sadly. I was at least able to find a new post about face recognition storage during my latest researches that I hadn't seen before.
With my limited stream programming knowledge, I can at least guess from Gal G's description that Gal did their configuration outside of the AcquireFrame loop, and before the loop occurs in the script, in order for that config to be done before the frame data is acquired.
Regarding creating a .dat file, I tracked down the posting of the Intel member involved.