Model Training results

100
0
03-12-2021 12:58 PM
Labels (2)
MartyRyan
New Contributor

First Model trained and need help with results - I'll add more for now, but posting these in hope of help getting a basic understanding of what they are telling me. Thanks 

Statistics: I exported 157 features for Deep Learning. Generated over 29000 image chips. Trained model and trying to read results.

MartyRyan_0-1615582678156.png

Learning Rate - slice(5.754399373371565e-05, 0.0005754399373371565, None)

Training Loss                                           Validation Loss                                       Accuracy

0.5108157396316528                             0.8722757697105408                             0.6979577541351318

0.5308706760406494                             1.2756025791168213                             0.5369244813919067

0.6114291548728943                             0.825732409954071                               0.7064075469970703

0.5830191373825073                             0.7337561845779419                             0.7257699370384216

0.5221974849700928                             0.8046979904174805                             0.7009243965148926

0.26255929470062256                           0.9559058547019958                             0.6519922614097595

0.3802303075790405                             0.8340374231338501                             0.7028971314430237

0.3805951178073883                             0.8353999853134155                             0.7071032524108887

0.3670295774936676                             0.7565853595733643                             0.7214756011962891

0.26321926712989807                           0.7860170006752014                             0.7285805344581604

{'accuracy': '7.2577e-01'}

NoData  Built Up  ...  Wetland Shrub  Barren Land

precision  0.190877  0.278278  ...            0.0     0.031376

recall     0.028441  0.360409  ...            0.0     0.011850

f1         0.036397  0.290547  ...            0.0     0.013370

[3 rows x 18 columns]

 

0 Kudos
0 Replies