Semi-autocomplete object detection annotations using semi-trained models
Currently only works with the Pascal VOC XML format using LabelImg and Keras RetinaNet.
Once annotations have been predicted, extraneous or innacurate annotations can be adjusted using LabelImg, which can then be used to get higher quality annotations at a much faster speed to continue training of original, semi-trained model.