Custom Yolo Model num classes mismatch

I am following the Custom YOLO Model guide when using deepstream 4.0 for Jetson Xavier and I am running into a num class mismatch when setting NUM_CLASSES_YOLO const variable in the nvdsparsebbox_Yolo.cpp file. This is the only place the guide is telling me to change when using my own YOLO models for the deepstream-app. I have also updated the config_infer.txt file with regards to the .cfg, .weights and label files as well as the num-detected-classes. Mu deepstream_config.txt is also pointing to the right config_infer.txt file as well as the same label file.

The mismatch seems to point out that the configured network correctly detects the classes but the detected network does not. Upon looking at the nvdsparsebbox_Yolo.cpp file, the detected network is based on the NUM_CLASSES_YOLO variable which i have already set. This suggests that I am updating the wrong cpp file to setup my own yolo to run in deepstream.

Am I missing anything with the setup to use my own YOLO model to test? Im not sure why the command deepstream-app -c deepstream_app_config_yoloV3.txt refers to a different bbox.cpp file.

Dear ralvarezl8d7,
It seems you are using Jetson Xavier. This forum is intended for DRIVE AGX platform. Could you please post your query at

Sorry my bad. I forgot to compile the code which is why changes did not reflect. We can close this.