OK, so you just use TLT classification network to train a resnet18 classifer.
In deepstream, I suggest you firstly check if this classifier can get the same accuracy as tlt-infer did.
as in, could you briefly explain that part. In deepstream we’d be running it on video. And on tlt-infer it’s on images. How should I compare?
Could you generate h264 file with some images and then use deepstream-test1-app to test against?
Or you can cross verify with this way:
dump your video clip into images frame by frame, and run tlt-infer to do inference. To check the results, especially the error frame you mentioned.
Hi beefshepherd,
We haven’t heard back from you in a couple weeks, so marking this topic closed.
Please open a new forum issue when you are ready and we’ll pick it up there.