GUI to label frames for training of ACM-dlcdetect
Project description
ACMtraingui
GUI to label frames for training of ACM-dlcdetect, by Arne Monsees
Installation
- Install Anaconda
- Start Anaconda Prompt (Windows) / terminal (linux) and navigate into repository directory
- Create conda environment
conda env create -f https://raw.githubusercontent.com/bbo-lab/ACM-traingui/master/environment.yml
Update
- Start Anaconda Prompt (Windows) / terminal (linux) and navigate into repository directory
- Update with
conda env update -f https://raw.githubusercontent.com/bbo-lab/ACM-traingui/master/environment.yml --prune
.
Running
- Start Anaconda Prompt (Windows) / terminal (linux) and navigate into repository directory
- Switch to environment
conda activate bbo_acm-traingui
- Run with
python -m ACMtraingui [options ...]
Options
Assistant mode
Run with python -m ACMtraingui [base data directory]
.
This starts a GUI in drone mode, for the use by assistants with limited options to influence how the program runs and were it saves. This expects the following file structure:
[base data directory]/data/users/{user1,user2,...}/labeling_gui_cfg.py
[base data directory]/users/
{user1,user2,...} will be presented in a selection dialog on startup. Marking results will be placed in [base data directory]/users/
Master mode
Run with python -m ACMtraingui [configdir] --master
.
This starts a GUI in master mode. Only do this if you know what you are doing.
Check mode
Run with python -m ACMtraingui [directory of labels.npz] --check [bbo_calibcam calibration npy]
.
This gives sorted text output of 3d and reprojections errors. Reporjection errors above 5-10px usually indicate errors in labeling and respective frames have to be checked.
Join mode
Run with python -m ACMtraingui [configdir of ACM-dlcdetect] --check [multiple directories containing labels.npz files]
.
This joins all marked labels in the labels.npz files into the labels.npz file in the dlcdetect configuration. Marked labels overwrite existing labels framewise.
TODO
- Document config
- Document sketch file (2d sketch of animal. If not presented, 3d wireframe is shown instead)
- Document model file
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for bbo_acm_traingui-0.10.1-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 6224dab7cec2021be04190ee6ea5c14b06e1a371403dcb80c1f9e84ffb74aa8d |
|
MD5 | 9b5751407fd4bd109752f8d5cdbeafd5 |
|
BLAKE2b-256 | 824a19edf2d12800b17f939c476c3c54e6b894a10ea04324a3c7202ad2bb690c |