Skip to content

Latest commit

 

History

History
95 lines (63 loc) · 2.39 KB

README.md

File metadata and controls

95 lines (63 loc) · 2.39 KB
Raw

benchmark / batch processing

How to run Pl@ntNet identification on a set of images, with or without a ground-truth to assess results quality.

usage

configuration

Edit the script before running it. You'll find a few adjustable options at the top of the code.

API_KEY

Add your API key here

mainFolderPath

Absolute path of the main data folder containing images subfolders (no trailing /); see "images folder" below

outputFilesPath

Folder where to place output files (no trailing /)

projects

Default value of [ "all" ] is acceptable for most cases. Add more projects if you wish to compare results: [ "all", "the-plant-list", "weurope" ] ; each image will be identified in every project)

resultsLimit

How many possible determinations will you keep in the output table, for every image (highest scores come first)

maxParallelQueries

How many HTTP requests will you make to the Pl@ntNet API at once. Max 10, defaults to 1 (sequential).

images folder

Organize your image files in subfolders named after ground truth as show below. Ground truth muth be a binomial name without author, separated by space or underscore. Case does not matter. Subfolders must contain image files in JPEG or PNG format. Image files don't have to be named after ground truth. If a subfolder contains another subfolder (name does not matter), all images contained in the latter will be processed at once, as a multi-image identification request.

main_data_folder
|
|__acacia_saligna
|  |__DSC00123.jpeg
|  |__DSC00124.jpeg
|  |__Folder with multiple images
|     |__DSC00145.jpeg
|     |__flower 1.png
|     |__DSC00158.jpeg
|
|__Acer campestre
|  |__Acer_campestre_L._2022-02-18.JPG
|  |__other maple pic.png
|
|__gunnera_tinctoria
|  |__whatever.jpg
|
|__IRIS GERMANICA
|  |__d8614efc6114b6a156a616514e16b151.jpeg
|
…

results

Run code (see below)

Results in CSV format are written to file ./plantnet_benchmark_{date-time}.csv

code

Node.js

Benchmark script using Node.js

install

npm install

run

nodejs benchmark.js

R

Benchmark script using R

R version does not support muti-images queries at the moment. Sub-folders inside species-names sub-folders are ignored.

R version is not parallelised at the moment.

install

install.packages("httr")
install.packages("wand")

run

Rscript benchmark.R