search

weecology / NeonTreeEvaluation

Benchmark dataset for tree detection for airborne RGB, Hyperspectral and LIDAR imagery
Creative Commons Zero v1.0 Universal
122 stars 22 forks source link

Version Control #37

Open WilliamLockeIV opened 1 month ago

WilliamLockeIV commented 1 month ago

Thanks very much for creating and making available this benchmark dataset! A quick question on version control: the Zenodo link on this Github repository leads to Data for the NeonTreeEvaluation Benchmark, for which the most recent version is 0.2.2 on Jan 27, 2022. However, two other Zenodo links we've found connected to the benchmark are weecology/NeonTreeEvaluation: New naming format, with version 1.8.0 on May 18, 2021, and Training Data for the NeonTreeEvaluation Benchmark with version 0.2.2 on Jan 27, 2022. The last of these is obviously training data, which is also included in "Data for the NeonTreeEvaluation Benchmark." So we wondered if the last dataset just contained training data, "weecology/NeonTreeEvaluation: New naming format" just contained evaluation data, and "Data for the NeonTreeEvaluation Benchmark" was a combination of the two? Or, if there are other differences, what dataset(s) you would consider most up-to-date for both training and evaluation for purposes of this benchmark? Thanks for the clarification!

bw4sz commented 1 month ago

Thanks for asking. There is some technical debt here as the repo is getting 5+ years old at this point.

https://zenodo.org/records/4770593

Should be considered the latest eval data.

Keep an eye on https://milliontrees.idtrees.org/, which will replace all of this with a proper python package and a very large benchmark.

What's your use case? Let me know if there is interest to contributing to the above effort.

WilliamLockeIV commented 1 month ago

Thanks very much! I'm working with a group on tree detection and crown segmentation from UAV imagery. I'd have to check with them to see if it's something we feel we could contribute to, but I appreciate knowing about it!