ResNet-18 on Digits problem

I am trying to use Digits to retrain the ResNet-18 (provided on the DeepStream SDK) on the kitti-dataset to detect Car and Van.

I edited the following parameters to generate the Object Detection database:
Pad image (Width x Height): (None)
Resize image (Width x Height): 640 368
Channel conversion: RGB
Custom classes: DontCare, Car, Van

And edited the following parameters for the Object Detection model:
Subtract Mean: None
Batch size: 6
Batch Accumulation: 4
Solver type: Adam
Base Learning Rate: 0.0001
Custom Network > Caffe
Custom Network: Copied the contents of “ResNet_18_threeClass_VGA_deploy_pruned.prototxt” as provided on the DeepStream SDK
Pretrained model(s): Path to “ResNet_18_threeClass_VGA_pruned.caffemodel”

The model is trained without any errors, however no information is displayed on the Loss/Epoch graph.

Then I use the Test tool on an image of the dataset and the following error is displayed:

The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

Could anyone please provide some insight about the meaning of this error and how to solve it?