TODO:

[sdtaylor]

manuscript:

• [x] do data/software citations
• [x] need a table/map of sites
• [x] need some example images
• [x] need license statement for all phenocam images
• [x] cite Borowiec 2021 - https://doi.org/10.32942/osf.io/nt3as
• [x] new figure 1
• [x] fix precision: dom cover in accuracy figs

rerun everything with:

• [x] sdtaylor/PhenocamCNN#6 sites excluded from the get go
• [x] other/fallow combined? Yes, only 86 images of fallow, 700+ of other
• [x] set the seed on held out data so they can be used later, or mark it in R somewhere
• [x] save val data frame for easier eval later

  Other things

• [x] deal with snow issue. If snow or water in dom category, then remove no crop in crop type and status just like removing blurry. Or actually changing to unknown sounds better.

  organize folders/files

  ie put A, B, C on folders for primary steps. prepend files with 1, 2, 3, etc.

[sdtaylor]

Get numbers/sample sizes for everything

• how many images in training set?
• how many in test set from fixed sites? from "filler" images?
• how many in final predictions timeseries?

[sdtaylor]

Zenodo repo

The git repo plus:

• [x] vgg16_v4_20epochs keras model file
• [x] vgg16_v4_20epochs keras model predictions
• [x] final_predictions.csv

[sdtaylor]

Revision TODO

• [x] cite MODIS Landcover HMM usage https://doi.org/10.1109/TGRS.2015.2463689
• [x] cite Liu et al 2021 https://www.sciencedirect.com/science/article/pii/S0308521X21002596
• [ ] highlight phenocam data in updated interpretive summary

[sdtaylor]

published