Centernet TFLite not detecting anything on Pixel 4a

General Discussion
, ,
#1

I have trained two ball detection models. One using SSD. And the other, Centernet. SSD works fine on my Pixel 4a (Android version 12) . But the Centernet model doesn’t. It gets initialised properly as well. It processes the image as well, but doesn’t give any inference. No crashes, errors or anything. However, the same centernet model works fine on my other phone, Samsung S9+.

Below, I have provided the centernet tflite file:

https: // drive.google .com /file/d/ 1qkYQo4Qa88AliZl7zv TQXx8C8uBb06vk/ view? usp=sharing

(spaces because i can’t post link here)

1 Like
#3

Hi Sohail,

It seems worrying that something is not working inside android 12. A lot of things have changed at this android SDK. Can you confirm also this problem with an emulator with android 12? Please get back here when you confirm…it is really interesting.

#5

Okay. I will give it a shot and get back to you. BTW, I also tested on another device which has Android version 11. It also doesn’t work there too.

#7

As I cannot use camera on an emulator (on my macbook), therefore I cannot test my application (model) on any emulator. is there any other way?

#9

Hi,

It seems that you have to upload somewhere the code and the files so I can check. Create a repository and share it here.

Thanks

#11

Sure.

By default, i have set the model to Centernet. Test it on an image with a football in it. And see the logs (logcat).
And if you want to change the model from CenterNet to SSD, Go to ModelManager → Line 26.

1 Like
#13

Hello! Any luck with this?