I have followed this guide to deploy yolov4 in a deepstream pipeline. I also compared it to the yolov3 guide and realised they are pretty much identical (1: Build TensorRT Open source software (OSS) 7.0) 2) Build the parsers (“YOLOv4 requires YOLOv3 custom bounding box parsers”)
Since I can already deploy a yolov3 .etlt model in my pipeline, I simply modified my deepsteam config to swicth to the corresponding yolov4 .etlt model. The yolov3 model detects objects correctly but the yolov4 detects nothing at all.
Why is that the case? Am I missing something?