Introduction to ReNom IMG ¶
ReNom IMG is model developing tool for object detection.
- New CNN architecture: YOLO version 2 is available.
- New Algorithm Setting: To train whole network or not can be selected.
- New feature: A page for creating dataset is added. Ratio of traingin data and validation data can be selected, and you can use them when creating new model.
- New feature: Preservation feature is added. At most 2 model can be train at the same time, and you can reserve model to train after finishing current training.
How to start
1.Type following command in ReNomIMG directory.
python -m renom_img
Or, following command is available from 0.7beta.
Second command can be called in any folder and creates datasrc folder at current directory.
Then ‘datasrc’ folder will be created in your current directory.
Please set images and labels according to 2.Set image data to ReNomIMG description.
How to use
The following videos describes how to use ReNomIMG.
In this video, the Oxford-IIIT Pet Dataset is used. - Cats and Dogs Classification
- Parkhi, A. Vedaldi, A. Zisserman, C. V. Jawahar
Cats and Dogs IEEE Conference on Computer Vision and Pattern Recognition, 2012 Bibtex
1.Create dataset directory
To create dataset directory, please start ReNomIMG server according to ‘How to start’.
The dataset directory is automatically created when you first start up server.
2.Set image data to ReNomIMG
As following video showing, please put the image and label you want to use in the datasrc/img and datasrc/label folder.
Also, please put the img you want to predict in the datasrc/prediction_set/img folder.
Same as before mentioned in ‘How to start’, following video describes how to start ReNomIMG.
4.How to crate datasets
Following video describes how to datasets.
5.How to start training in ReNomIMG
Following video describes how to start model training in ReNomIMG.
6.How to use model for prediction.
Following video describes how to use model for prediction in ReNomIMG.
Format of xml file.
The format of the xml file which created by ReNomTAG follows PASCAL VOC format.
- OS: Ubuntu 16.04
- Browser: Google Chrome(version 63.0.3239.132)
- Python: >=3.5