Overview
Although DORA offers extensive customization, the DORA team has established default settings as a starting point.
Your input data and targeting goals may require deviations from these defaults, and the validation graphs generated during the process will likely prompt adjustments to various parameters across Steps 1-6.
Read on to review the default settings on a step-by-step basis.
Default Parameters by Section
AOI Selection
Height (px): 512
Minimum Value: 256
Maximum Value: 4096
Width (px): 512
Minimum Value: 256
Maximum Value: 4096
Input Features
Features: there are no default values. Select the files to include in the experiment.
Correlation threshold: 0.90
Minimum Value: 0.00
Maximum Value: 1.00
Set Up Learning Data
No defaults: this is where you set targets, which are dependent on your own files and preferences.
Embed Visual Features
Vision Transformer Model: no default. Pick the most relevant Vision Transformer Model for your targeted commodity or commodities. When in doubt, use Master_Model.pt.
No. of epochs: 100
Minimum Value: 10
Maximum Value: 1000
Build Predictive Models
Minimum Cluster Size: 150
Minimum Value: 10
Maximum Value: 1000
Minimum Samples: 10
Minimum Value: 5
Maximum Value: 100
Advanced
Number Simulations: 10
Minimum Value: 1
Maximum Value: 30
Neg Balance: 1.20
Minimum Value: 1.00
Maximum Value: 3.00
Tree Depth: 24
Minimum Value: 12
Maximum Value: 192
Nb Trees: 300
Minimum Value: 50
Maximum Value: 1000
Min Split: 20
Minimum Value: 1
Maximum Value: 100
Predict depth: Yes
Generate Targets
Target Threshold: 0.50
Minimum Value: 0.05
Maximum Value: 1.00
Maximum Distance: 100
Minimum Value: 10
Maximum Value: 1000
Minimum Samples: 10
Minimum Value: 1
Maximum Value: 100
Notes
The settings for each step will vary widely depending on a variety of factors like your targeted commodity, quality and quantity of your input data set, the other settings, and the results.
Still have questions?
Reach out to your dedicated DORA Contact or email Support@VRIFY.com for more information.