Unet HLD model results

[muthukumaranR]

initial Model Summary: 3 layer standard Unet

• Total params: 175,313
• binary crossentropy
• adam optimizer
• metric: accuracy

Run #1: precision: 0.79 recall: 0.70 accuracy: 0.83, f1_score: 0.749

[muthukumaranR]

naturally, The snapshot of the model with the lowest validation loss is stored for evaluation and further use.

[muthukumaranR]

model overfitting, reducing param # and re-running

[muthukumaranR]

Some sample predictions on test set:

[muthukumaranR]

since The predictions look good, the model just needs to be trained with a lower learning rate for a smoother graph. It just learns too fast, especially with the augmentations included.

[xhagrg]

@muthukumaranR are we getting these results from the same model posted above the images?

[muthukumaranR]

yes, @xhagrg

[xhagrg]

hmm, let's try with some other test samples for this. see if we get better/worse results. (skeptical because of the graph)

[muthukumaranR]

predictions can be found here:https://drive.google.com/drive/u/1/folders/15FfMvlFUL9qajKrbrim_bhOfjrl3Ucld

[xhagrg]

any false positives? let's run it for a day's worth of data.

[muthukumaranR]

Updated loss plot with a slower learning rate

[xhagrg]

@muthukumaranR noice!

[muthukumaranR]

Testing the model on random NON-HLD images for a day's worth of data from worldview, the model is susceptible to false positives. Taking these example and adding them to the training/validation cases to reduce false positives.

[muthukumaranR]

some false positive cases from the model:

[muthukumaranR]

The false positives were due to not scaling the images. once the images are scaled, there were no false positives.