I have one feature that I am trying to extract from all my images (UAV imagery at 0.04m pixel). I have approximately 150 training samples of that feature from which 81 chips have been created (though 2 are not right). However, when I Train the Deep Learning Model I get empty performance values for each Epoch, and when I look at the model_metrics the graphics are empty, with just the chip images side by side.
When I try to run the Detect Objects on the original training raster, I cannot select my DLPK stored locally, after copying the folder from a server (where it was originally created).
Am I missing a very basic step here?
Thanks
Train Deep Learning Model
=====================
Parameters
Input Training Data G:\Users\jw75kg\Madagascar\Ankarafantiska_2022_2024_GuardianAppeal\Yam\UAV_Assessment\yam_holes_2022_Pachy_Train
Output Folder G:\Users\jw75kg\Madagascar\Ankarafantiska_2022_2024_GuardianAppeal\Yam\UAV_Assessment\yam_holes_2022_Pachy_Train_DLmodel
Max Epochs 20
Model Type FEATURE_CLASSIFIER
Batch Size 64
Model Arguments
Learning Rate
Backbone Model RESNET34
Pre-trained Model
Validation % 10
Stop when model stops improving STOP_TRAINING
Output Model G:yam_holes_2022_Pachy_Train_DLmodel.dlpk
Freeze Model FREEZE_MODEL
Data Augmentation DEFAULT
Augmentation Parameters
Chip Size 256
Resize To
Weight Initialization Scheme
Monitor Metric VALID_LOSS
=====================
Environments
Processor Type GPU
=====================
Messages
Start Time: 07 June 2024 16:36:46
Learning Rate - slice(2.51188643150958e-08, 2.51188643150958e-07, None)
epoch training loss validation loss accuracy multi_label_fbeta time
0 0.0 NA 00:00:03
1 0.0 NA 00:00:02
2 0.0 NA 00:00:02
3 0.0 NA 00:00:02
4 0.0 NA 00:00:02
5 0.0 NA 00:00:02
6 0.0 NA 00:00:02
{}
Succeeded at 07 June 2024 16:43:22 (Elapsed Time: 6 minutes 35 seconds)
