Inference over Facenet makes different results on Jetson Nano and dGPU

[IsraelLencina]

Hi, i'm using this repo from a laptop with TensorRT (it's TensorRT 20.03 docker image), and also from a Jetson, the code works without problems in Jetson, the detections are OK this is in expected values.
But when i execute this repo on my own laptop, it return values between 500 and 1000 do you know why is it?

[shubham-shahh]

Hi, i'm using this repo from a laptop with TensorRT (it's TensorRT 20.03 docker image), and also from a Jetson, the code works without problems in Jetson, the detections are OK this is in expected values.
But when i execute this repo on my own laptop, it return values between 500 and 1000 do you know why is it?

Hi, 500 and 1000 are values of what? kindly eleborate.

[imSrbh]

@shubham-shahh I think these are the embeddings.

@IsraelLencina I am also trying to reproduce this setup on dgpu, facing saveral issues.

• h5 to pb conversion

[IsraelLencina]

Yes, it's embeddings, i've seen that the change is after the inference.

[IsraelLencina]

Hi again, i've seen in NOTES section that .uff and .engine are GPU specific, i've tried to generate everything from step 3 (as you say in NOTES section) but the "GPU embeddings" are still in values bigger than "Jetson embeddings".

[shubham-shahh]

Hi again, i've seen in NOTES section that .uff and .engine are GPU specific, i've tried to generate everything from step 3 (as you say in NOTES section) but the "GPU embeddings" are still in values bigger than "Jetson embeddings".

are you referring to the master branch or develop branch?

[IsraelLencina]

Always master