This repository hosts training data labeling and related program, excluding collected training data. The whole project with collected training data can be downloaded here (it's huge).
You may create your own labeling utility programs, as long as the output result.csv
meets the same format.
Currently, here's how we organize and label collected data:
- Create a folder named after the collecting date. Put in images(collected by PlantsCam-iOS) with iTunes File Sharing.
- Run
./walk
to createresult.csv
. Afterwards, simply run./walk -a xxxxxxxx
to add the new folder xxxxxxxx to append toresult.csv
. (You need to rungo build walk.go
before using it) - Label the data in
result.csv
and commit it.
Plus, run ./total
to see total classes labeled. When new classes are added, run ./total -o xxxxxxxx
to create or update the doc. See more in doc/ directory.
See the project plant-tw/models_plant
The program is licensed under GNU Affero General Public License v3.0. Check the LICENSE file.
The collected training data are licensed separately. Check the LICENSE file in each project folder.