Experiment with augmentation more
robmarkcole opened this issue · 1 comments
robmarkcole commented
Getting pretty unsatisfactory performance. The article below highlights that flames are a difficult target: wildfires have one of the most complicated features to obtain through CV, including flame height, flame inclination angle, firebase width, etc. Any deviation or improper prediction of the measures above can result in poor model performance.
Image augmentation should help
robmarkcole commented
- In ice-vision notebooks we are applying default augmentations with
aug_tfms