In the seventh blog, when our code running. We should write a number from 0 until 9 to see recognition of the AI. after write the number, close the white board and the code will execute the number. It will tell us the predict number and there is a question “is this correct number ?” to make sure the number correct or no. after that, we should choose the font text to see comparison between our writing and font text.

This project kindly hard to code because it needs a lot of libraries to run this code. Last time, we confused to use the code and really complicated. Also, in the pycharm could not install all libraries. We decided to use anaconda to download several libraries and used several github as our references.

In the fourth blog, we did the code testing. We tried to run the code with different images of writing as the input. It also to understand the workflow. We also try to find another referenced in the youtube. It is talking about how to use Keras and TensorFlow codes at the same time and the library uses mnist for the number zero until nine in large twenty-eight times twenty-eighth. Finally, we decide to use from github to our reference is https://github.com/githubharald/SimpleHTR

In the fifth blog, Our project uses three algorithms to recognize handwriting which are Convolutional neural networks (CNN), recurrent neural networks (RNN)  and a Connectionist Temporal Classification. CNN algorithm used to identify the relevant parts of the fed image. The output will downsize the image. RNN will identify the relevant information of the sequence from the output. At the same time when RNN running will output matrix. It used for CTC which will use the text on the ground to calculate the value of the loss.

In the sixth blog, our code using TensorFlow that implemented in the code library to read and convert images into text. Our code provided by Niek Temme is the basis of our handwriting algorithm. We used MNist database for our project which has a 28×28 scale. When the code running it will appear a square white to writing. After that, the writing will become an image and will be compared to those in the MNist database. Also, Predint() is the main function to produce the prediction based on the image.

In the third blog, we tried to run the code. However, the code was not running so well. The code needs environment from TensorFlow. On the first tried we just install the TensorFlow from pycharms libraries, but it could not install. Tensorflow has a compatibility issue with the latest version of Python 3 and we need to downgrade python 3 environments. After that, we found the other way to get TensorFlow from Anaconda. So we download anaconda first and we used the anaconda prompt to install TensorFlow environments:

1. conda create -n tensorflow_env python=3.6
2. Pip install tensorflow==1.12
3. Pip install opencv-python
4. Pip install editdistance
5. Pip install matplotlib

In the second blog, our group already decided to choose Python 3 for coding the language in this project. In python, almost all directory already prepares or easy to install. Furthermore, our group does not really understand what a computer vision is. Because of that, our group read several websites to understand individually on it. After we read it on the internet, we found code from Harald ScheidI’s Github. It uses python and needs to install the Tensorflow v1.12 library. It is a primary basis to read the handwriting.

17 until 23 October our group decided on the idea for the project, it is about handwriting. Our group focuses on this because the majority of people already know about technologies. It will help other people more easily to write a word digitally and also for inserting a picture. Our team chooses to handwrite for the project because we wanted to make something different from the other groups. Commonly, many Intelligent systems used for the games to make an AI. We decided on this project to make people can write on digital text to compare a certain font and fix their own handwriting. Our team hopes handwriting project would be useful for people to attempt their writing with appropriate settings.

Welcome to Binusian blog.
This is the first post of any blog.binusian.org member blog. Edit or delete it, then start blogging!
Happy Blogging 🙂