I am trying to create a script (relatively new to python) to automate running ExportTrainingTool, but I keep getting an error. I have all the necessary packages installed as far as I can tell and I'm in my "deeplearning" environment where I added the packages. The error I'm assuming has to do with the backbone model parameter and the error I get is:
NameError: name '_resnet_family' is not defined
I have image analyst, and I have previously run this as a gp tool in ArcPro successfully (a few months ago so something could have changed).
My parameters are:
in_folder = "C:/Users/18602/Documents/Masters Program/final_project/Geog6180_FinalProject/LabeledObjects"
out_folder = "C:/Users/18602/Documents/Masters Program/final_project/Geog6180_FinalProject"
max_epochs = 25 #A maximum epoch of one means the dataset will be passed forward and backward through the neural network one time.
model_type = "SSD" #Single Shot Detection - object detection. This is the method of the model, and was chosen as it is the example method for well pad selection
batch_size = 2
arg = "grids '[4, 2, 1]';zooms '[0.7, 1.0, 1.3]';ratios '[[1, 1], [1, 0.5], [0.5, 1]]'" #default grid zoom and ratio values
#learning_rate = 0.003
backbone_model = "RESNET34"
#pretrained_model =
validation_percent = 10 #default
stop_training = "STOP_TRAINING"
freeze = "FREEZE_MODEL"
# Execute
TrainDeepLearningModel(in_folder, out_folder, max_epochs, model_type,
batch_size, arg, None, backbone_model,None,
validation_percent, stop_training, freeze)