I am trying to develop an app to render a 3D model of a watch on my wrist. But the Sceneform SDK detects a plane only.
I tried augmented images by augmenting the picture of my wrist, but my wrist was not detected due to too few feature points.
Any suggestion on how it can be done? I am an Android Developer and use Sceneform SDK with ARCore. I have no experience with Unity.
Please help!!
You need to use a Google ML Kit model along with ARCore and Sceneform. This Machine Learning framework allows you to add to your app such features as Faces/Landmark Detection, OCR, Image Labelling, QRcode/Barcode scanning, etc.
AI can detect your wrist using Custom Model Inference (then ARCore can place an Anchor on it). But you have to implement a TensorFlow Lite pre-trained model.
ML Kit is for Android and iOS.
Hope this helps.
While I love the idea of using ML Kit to build your own computer vision model, that's probably a tad more involved than you were hoping to make things.
Might I suggest that a quick and easy solution would be to stick a sticker on your wrist that is registered as an image target with ARCore? You could even just draw a marker on your wrist with some success.
You could then make some assumptions about the size of an average wrist for setting up some basic occlusion effect, in order to hide the bands of the watch on the other side of your wrist.
