Detection of images using Object Detection Algorithm based on YOLO (importing tensorflow) in different setups

[AnishKumarNayak]

Hi,

I have been running some detections of images by calling upon trained weights for YOLOv3. This is all done through either tensorflow or tensorflow-GPU.
Some of these detections are done through my own laptop (CPU bases -- Mac) and sometimes using GPU Clusters which we can only access by SSH.
I noticed that for the very same weights being used, on different laptops and CPU workstations the detections are the same.
However, in case of using a GPU cluster or CPU cluster which does not have any display input, the detections are different.
Apart from this, the workspace is identical.

Does this have to do anything with the fact that some display drivers are not installed or some tensorflow libraries are missing??

I know this is a pretty generic query, and I'll try my best to elaborate further if required.

Thanks

[bileschi]

Hi @AnishKumarNayak
This is probably a better request for TensorFlow (the ML Framework) than TensorBoard (The analytics and visualizations system). Please re-open if I am in correct (sorry) and there is evidence that this is actually a TensorBoard issue.

[AnishKumarNayak]

My bad, I had also thought that I submitted in the Tensorflow main thread. Thanks for pointing that out.