The OCDNet model detects optical characters or text on images.
The training algorithm inserts the binarization operations into the segmentation network and jointly optimizes them so that the network can learn to separate foreground and background pixels more effectively. This model was trained using the OCDNet in TAO Toolkit.
The primary use case for this model is detection against characters or text on images.
The datasheet for the model is captured in its model card hosted at NGC.