how to cat left_rectified and right_rectified in a pipeline

ynjiun

I built a custom model which will take a combined_img as input. The combined_img is defined as following:

def preprocess(left_rectified,right_rectified):
    left_img = np.expand_dims(left_rectified, 2)
    right_img = np.expand_dims(right_rectified, 2)
    combined_img = tf.convert_to_tensor(np.concatenate((left_img, right_img), axis=-1) / 255.0, dtype=tf.float16)
    return combined_img

Is there a way to build the following pipeline using depthai?

left_rectified
        |
         ------->  preprocess(left_rectified,right_rectified)  -->  custom model blob 
        |
 right_rectified

Could you please help to point me to a reference code or example links to show how to build the above pipeline and run in oak camera?

Thanks a lot for your help.

erik

Hello ynjiun ,
For creating such custom models I would suggest checking our docs here, and example for pointcloud conversion. We also have some demos here. Specifically, you should check out the concat example, which is essentially what you are looking for.
Thoughts?
Thanks, Erik