Benjamin Weinstein

[WIP] NEON species benchmark data

11

Hi everyone. I had a brief chat with @calebrob6 and others about including our data here. We (@ethanwhite and @MarconiS and others) are developers of the NEON data science competition...

Comparison to Zooniverse, track users?

4

This looks like a great tool! Well done. We'd be interested in setting up an instance for [Tree Detection](https://www.biorxiv.org/content/10.1101/790071v1.abstract) (also a AI4earth project). Couple quick questions as I scan the...

DOC: Can I perform operations on the geocube pixels during rasterization? Count, Mean, etc.

11

Hi this is my first foray into geocube. Thanks for your work here, clearly a good idea. I'd like to add an example usage that I believe should be possible,...

US census data

1

# key aspect population, race, ethnicity, age, by county. might already exist: https://github.com/walkerke/tidycensus although it might be useful in using that API to pull in data

2014 survey of the tree species and location at Harvard Forest, one of the largest plots in the country. 2020 data is expected soon. https://harvardforest1.fas.harvard.edu/exist/apps/datasets/showData.html?id=hf253

Both spatial .shp and .csv files. https://zenodo.org/record/3765872

NEON Tree Dataset

4

The woody vegetation structure data from NEON. confirming with @MarconiS The date of release is key, since data collection is ongoing. https://github.com/weecology/DeepTreeAttention/blob/master/data/raw/neon_vst_data_2021.csv The components and column definitions from this dataset...

One of the largest ecological plant-pollinator networks collected. Released in this paper: https://onlinelibrary.wiley.com/doi/abs/10.1111/ele.12730 Data are here: https://github.com/bw4sz/BetaNetwork/tree/master/data Key dataset is https://github.com/bw4sz/BetaNetwork/blob/master/data/HummingbirdInteractions.csv Key Columns: "ID": Camera placement "Video": Video ID "Date":...

Model should be less sensitive to resolution

6

Overall the models are too sensitive to input resolution. There are two strategies to improve, 1) the preprocessing, 2) the model weights. # Preprocessing As the resolution of the input...

Ask user if they want to download the latest model.

8

We have not specifically gotten into this problem yet, but there should be a mechanism to ask user if they want to download a new model, instead of forcing it...