Shape must be rank $ but is rank 3 for 'model/MirrorPad'

CVLAB-Unibo / Real-time-self-adaptive-deep-stereo

Code for "Real-time self-adaptive deep stereo" - CVPR 2019 (ORAL)

Apache License 2.0

420 stars 73 forks source link

Shape must be rank $ but is rank 3 for 'model/MirrorPad' #74

Open EhrazImam opened 3 years ago

EhrazImam commented 3 years ago

I am using 2 IMX219 camera in Jetson Xavier NX when i am trying to run demo_model.py............ In grabber.py i have added my custom camera code tooo which i have already tried it and its working fine.... but i am getting error as you can see below... WhatsApp Image 2021-09-14 at 9 19 57 PM

AlessioTonioni commented 3 years ago

Try to add a dimension to the image tensor, each input frame should have shape [1, 480, 640, 3] in your case. You seem to be missing the 1 in the first dimension.

EhrazImam commented 3 years ago

I have tried that too... WhatsApp Image 2021-09-15 at 1 34 33 PM

AlessioTonioni commented 3 years ago

The error is a different one now. You are feeding an image with shape 360 x 640 to a placeholder with shape 480 x 640. Either resize your image to the target shape or change the shape of the placeholder.