HDLNet: design and development of hybrid deep learning network for optimally recognising the handwritten Kannada characters

At first, the Kannada character images are collected via benchmark datasets. After image collection, it is undergone the feature extraction process. Here, the extraction techniques are employed to acquire geometric features, texture features, and morphological features. Further, it is fused together with an optimal selection of features with optimal weights, thus it is provided as weighted fused attributes. Here, the optimisation of weight is done by the developed Fish-based Position of Marine Predators and Forest Optimisation (FP-MPFO). At last, the features which are weighted are given to a Hybrid Deep Learning Network (HDLNet), where the two models like Dense Long-Short Memory (DLSTM) and Attention-Based Deep Temporal Convolution Network (ADTCN) are incorporated with each other. To acquire the optimal value, several parameters are optimally tuned by developed FP-MPFO. Hence, the key outcomes illustrate that it has the potential to recognise the Kannada characters effectively. ©, Engineers Australia.