Run on external image/video

[fabiozappo]

Hi thank you really much for your work on this Pytorch version!

I would like to run the code on a single external image/video. Does your code provide a demo as the original one?

[rabysam28]

@varunagrawal @rogerhcheng I wanted to follow up on this ticket, since I also am interested in just testing one single external image to see how it works for my cases. Is there a python demo script for this? Thanks

[varunagrawal]

I can add a python demo script next week. Sorry I don't get issue notifications from github for some reason.

[gdiasbruno]

Hey @varunagrawal ,

First of all, thanks a lot for your amazing work! It will help me in my master thesis

Do you have a script? If you do, amazing, but I see it is also possible to create one from the evaluation script.

Thanks a lot, again!

[varunagrawal]

The trainer.py module has a function called get_detections which is what you're looking for.

I am in the process of refactoring the codebase so expect things to change. I intend to add a more straightforward means of deploying the model.

[varunagrawal]

@fabiozappo see the evaluate.py script. It evaluates on the test dataset, but you can see that the get_detections method takes an image tensor as input.

Closing this issue.