Does Spatial SDK Support Custom Object Recognition?
Does the Meta Quest Spatial SDK support custom object recognition or ML model retraining, or does it only use Meta's pretrained models?
Forum Discussion
1 Reply
Hi!
Using the Passthrough Camera Access API, you can implement custom object recognition with any model you like.
The Spatial Scanner app was built with Meta Spatial SDK, and it uses ML Kit for model inference and efficientnet TFLite 4 uint8 for the model.
Documentation for the app can be found here .
An in-depth breakdown of why these tools were chosen, can be found here.
