Run a dummy image to generate the model variables

I have been stuck on the below error for days now. Not been able to figure out the reason, I have already completed all the code passing all the tests (though the losses do not converge neither do I see any bounding boxes in the end).

When I run the above cell for dummy image, multiple times, it passes saying ‘Weights restored!’, however the shape from the next cell’s output comes out to be (3, 3, 256, 24)
(256,)
(256,)

which should have been
(3, 3, 256, 24)
(512,)
(256,)

Any help would be really appreciated as this is the only assignment that I need to pass for the whole course.

Hi,

This is a bit of complex assignment, I would suggest you check carefully how you define the checkpoint layers you are using from the model, there should be a wrong one chosen. Go through the previous steps again.

Hi, Thanks for your reply.

I can’t find any error in the checkpoint layers as the outputs matched the Expected Output for all the cells. I used tf.train.Checkpoint, restored the checkpoints the way it should be. The last output for checkpoints was -
“<tensorflow.python.training.tracking.util.CheckpointLoadStatus at 0x7f42b3097c90>”

I can share my colab link. Kindly advise.

I was able to figure out the error which was due to the wrong pipeline config path. It was pointing to SSD resnet 152 instead of SSD resnet 50 640x640. corrected it and passed the assignment. Didn’t have to modify any other cell, just the pipeline config path.

1 Like