AlbertoSabater
commented
1 year ago Hello. You can try to create an empty scores vector with just the score of the retrieved class_id, the remaining class scores could be 0.0. However, since REPP has not been designed to work in that way I cannot know if the final performance will be a good one. Let me know how it works in that scenario if you try.
I use tensorrt model to infer images，which output class score with its class id。If I don‘t have class scores vector ，can repp work？