ML model is based on Keras implementation of Convolutional Recurrent Neural Network for text recognition. The CRNNs are the combination of two of the most prominent neural networks. They involve CNN(convolutional neural network) followed by the RNN(Recurrent neural networks). The proposed network is similar to the CRNN but generates better or optimal results especially towards audio signal processing and text recognition.
In our example model consists of Conv2D layers and unidirectional LSTM layers.
On screen there is a horizontal RecyclerView that displays colored images with text. The background and the text are of different random colors. User picks on of the images and then 2 procedures start. First, the image is converted to grayscale, then one of the channels is isolated to generate the bytebuffer and lastly the flex delegate performs inference. Second, ML Kit's Textrecognition is used to perform OCR on the same grayscale image. When the two procedures stop user watch on screen the output and the total inference time of the procedures.