Error with Compreface component defined #403
-
Hello, I'm testing Viseron 2.0b3 on a Jetson Nano with Jetpack 4.6. Decided to try enabling Compreface face recognition. During Viserson start, this error snippet is displayed
This is the compreface section within config.yaml
Additionally, does Viseron not use any prexisting face collections in Compreface? The component doc states that faces are loaded from the face_recognition_path and a model is built. Thanks and regards. |
Beta Was this translation helpful? Give feedback.
Replies: 5 comments 18 replies
-
Looking into it, running some new builds as we speak but it takes some time for them to complete |
Beta Was this translation helpful? Give feedback.
-
Should be fixed now if you pull the latest version of Please try it out when you get some time and let me know how it goes.
You need to tell Viseron what people to track using the face_recognition_path, yes. However you can just opt to set |
Beta Was this translation helpful? Give feedback.
-
The latest dev build starts up without errors when Compreface is enabled. Thank you for responding so quickly and for clarifying the role of existing collections. |
Beta Was this translation helpful? Give feedback.
-
I noticed that the following error pops up intermittently and randomly for the 2 cameras that I'm testing with Compreface face recognition. It could occur singly or in bursts. I can't discern a pattern as to when or how. Not sure where to dig further.
This is with current |
Beta Was this translation helpful? Give feedback.
-
I'm happy to report that after 21+ hours with with current dev build, the error has not appeared at all . Thanks! Very intermittently, some uncaught exceptions pop up - usually relating to "concat_segments". Should I report those? |
Beta Was this translation helpful? Give feedback.
Should be fixed now if you pull the latest version of
roflcoopter/jetson-nano-viseron:dev
Please try it out when you get some time and let me know how it goes.
You need to tell Viseron what people to track using the face_recognition_path, yes. However you can just opt to set
train: false
and use your already trained model, but you still need the folder structure for Viseron to pick up the names of people.Now that you mention it, it sounds kind of weird, would be a lot smoother if it would fetch the trained …