Implement livestream image classification

[ghost]

No description provided.

[PaulTR]

Hey Duy, would you mind updating the readme to explain what this app does, how to run the model download script (and that it needs to happen), and maybe a little information about how the code is set up/key classes.

Thanks!

[PaulTR]

Alright should just need to run dart format. I also ran into an issue where the camera froze and crashed the app when running from visual studio to Android, but it works fine in Android Studio, so I'm guessing it's more of a setup issue on my side.

[PaulTR]

Can you also update the UI to add a small section at the bottom to show the classifications (normalized as 0.0 -> 1.0) that is more easily visible (like what we have in the Android apps). We can keep it sorted and displaying 3 items consistently rather than changing those options at runtime. Thanks Duy!

[tenkeyless]

Can you check for memory leak?

I think it is because if inference.run method on tflite.
I've run classification with your code, but I used my own mobileNetV2 with float16.

When I run 6 minute, memory goes up to about 630MB.

When I run 9 minute, memory goes up to about 700MB.

When I run 43 minute, memory goes up to about 2.09GB.

Memory goes up and never go down, even I change it to gallery mode.

I ran that with iPad pro 11 m1 model.

[PaulTR]

Memory issue on iOS being tracked in #110 and outside of the control of the sample app.

[tenkeyless]

After your comment, I've found out it is outside of the control of the sample app. It happened only iOS.

[PaulTR]

LGTM. Thanks Duy!