BoxAL is an active learning framework that automatically selects the most-informative images for training an object detector (like Fast R-CNN or Faster R-CNN) in Detectron2. By using BoxAL, it is possible to reduce the number of image annotations, without negatively affecting the performance of the object detector. Generally speaking, BoxAL involves the following steps:

1. Train an object detector on a small initial subset of a bigger dataset
2. Use the trained object detector to make predictions on the unlabelled images of the remaining dataset
3. Select the most-informative images with a sampling algorithm
4. Annotate the most-informative images, and then retrain the object detector on the most informative-images
5. Repeat step 2-4 for a specified number of sampling iterations
   
   

Linux/Ubuntu: INSTALL.md

Windows: INSTALL_Windows.md

Split the dataset in a training set, validation set and a test set. It is not required to annotate every image in the training set, because BoxAL will select the most-informative images automatically.

1. From the training set, a smaller initial dataset is randomly sampled (the dataset size can be specified in the boxal.yaml file). The images that do not have an annotation are placed in the annotate subfolder inside the image folder. You first need to annotate these images with LabelMe (json), V7-Darwin (json), Supervisely (json) or CVAT (xml) (when using CVAT, export the annotations to LabelMe 3.0 format).
2. Step 1 is repeated for the validation set and the test set (the file locations can be specified in the boxal.yaml file).
3. After the first training iteration, the sampling algorithm selects the most-informative images (its size can be specified in the boxal.yaml file).
4. The most-informative images that don't have an annotation, are placed in the annotate subfolder. Annotate these images with LabelMe (json), V7-Darwin (json), Supervisely (json) or CVAT (xml) (when using CVAT, export the annotations to LabelMe 3.0 format).
5. OPTIONAL: it is possible to use the trained model to auto-annotate the unlabelled images to further reduce annotation time. Set auto_annotate to True in the boxal.yaml file, and specify the export_format (currently supported formats: 'labelme', 'cvat', 'darwin', 'supervisely').
6. Step 3-5 are repeated for several training iterations. The number of iterations (loops) can be specified in the boxal.yaml file.

Please note that BoxAL does not work with the default COCO json-files of detectron2. These json-files contain all annotations that are completed before the training starts. Because BoxAL involves an iterative train and annotation procedure, the default COCO json-files lack the desired format.

Open a terminal (Ctrl+Alt+T):

(base) user@computer:~$ cd boxal
(base) user@computer:~/boxal$ conda activate boxal
(boxal) user@computer:~/boxal$ python boxal.py --config boxal.yaml

Description of the other settings in the boxal.yaml file: MISC_SETTINGS.md

Please refer to the folder active_learning/config for more setting-files.

Use a trained object detector to auto-annotate unlabelled images: auto_annotate.py

Example syntax (auto_annotate.py):

python auto_annotate.py --img_dir datasets/train --network_config COCO-Detection/faster_rcnn_X_101_32x8d_FPN_3x.yaml --classes healthy damaged matured cateye headrot --conf_thres 0.5 --nms_thres 0.2 --weights_file weights/broccoli/model_final.pth --export_format labelme

See TROUBLESHOOTING.md

Our software was forked from Detectron2 (https://github.com/facebookresearch/detectron2). As such, the software will be released under the Apache 2.0 license.

Please have a look at our active learning software for Mask R-CNN (which formed the basis of BoxAL):
https://github.com/pieterblok/maskal

BoxAL is developed and maintained by Pieter Blok.