How to train detectnetv2 on darknet architecture?

I am trying to train my custom dataset on detectnetv2 with darknet 19 architecture but I could not find correct configuration of parameters My configuration file is
detectnet_v2_train_darknet19_kitti.txt (3.0 KB)
.
I am running this command but getting error of shape:

detectnet_v2 train -e $SPECS_DIR/detectnet_v2_train_darknet19_kitti.txt \
                        -r $USER_EXPERIMENT_DIR/experiment_dir_unpruned \
                        -k $KEY \
                        -n darknet19_detector \
                        --gpus $NUM_GPUS

.

If you run with 3.0_dp_py3 docker, please make sure you have resized all of your images/labels offline to match your training spec(1280x720).
If you run with 3.0_py3 docker, please refer to below. See DetectNet_v2 — Transfer Learning Toolkit 3.0 documentation

The train tool does not support training on images of multiple resolutions. However, the dataloader does support resizing images to the input resolution defined in the specification file. This can be enabled by setting the enable_auto_resize parameter to true in the augmentation_config module of the spec file.

I am using the images of same resolution. My issue was due to wrong parameters in txt file .
I finally started the training using this spec file
detectnet_v2_train_darknet19_kitti.txt (3.1 KB)
Thanks for your help.

1 Like