Intelligent system week 4 progress

9 dec – 16 dec

After doing some research we found out that three neural network layer can be used to recognize handwritten letter.The CNN(convolutional neural network) layer, as the first layer is used to catch the main feature of the letter image which will be downsized along with feature map being added. After the image is processed in CNN, the image’s output is processed further in RNN, in RNN the  sequence is read to get information reagarding the relevant part of image. To improve the accuracy of finding the LSTM(longger Short Term Memory) can be used to generate information through longer distance which subsequently imrpve the finding accuracy. The process in rnn will result in matrix which will be passed to CTC, where the final computation will be done. the computation in CTC includes, computation of loss value using the ground  truth test, decoding of the final test. 

Leave a Reply

Your email address will not be published. Required fields are marked *