testing-model.md

↤ BACK

Testing the Model

Once training has started on our model, we can start testing it:

Model Validation

Generate Analytics on Model

YOLOv8 has some powerful validation tools which will generate a bunch of handy charts and graphs.

python validate.py

This will create a new folder in ./runs/segment/val ( each time you run this, a new folder is created, e.g. val2 etc ).

Model Predictions

This is the fun part - Testing that the model works.

We have a default samples folder which will be used with a confidence of 0.01.

python predict.py nano

Size Options: nano | small | medium | large | extralarge

This will create a new folder in ./predictions/fbc-seg-{size}-e{epoch} where {size} is the first letter of the size, and {epoch} is the number of the last epoch that ran

CLI Options:

--confidence

You can adjust the confidence of the prediction by passing in a --confidence of -c flag ( 0.0 - 1.0 ):

• Lower confidence has more predictions returned, but less accurate
• Higher confidence has less predictions returned, but more accurate

python predict.py --confidence 0.25

--image

You can pass in any image you want to test. We recommend using an absolute path.

python predict.py --image /path/to/image.jpg

---