You are required to train a neural net model for the handwritten data [login to view URL] script will be used to evaluate each code.
Only use python version 3 (not 2) for coding.
Input to file:
Your code should accept command line parameters. It should take one image file name as
python3 [login to view URL] [login to view URL]
Input image file will have fixed 20x20 size in grayscale (similar to the one displayed in the
class exercise). In the dataset images are saved in the form of rows of 400 length.
These rows can be converted into 20x20 grayscale images by using following python code.
(You may import packages form numpy and matplotlib)
[login to view URL](filename, [login to view URL](data).reshape(20,20).T,cmap="gray")
In the example above data is the row data of X. for example X, X so on, “filename”
is the name of the image file. It can be any name like “[login to view URL]”, “[login to view URL]” so on.
Output of file:
Output of your code should be the predicted digit. Only digit in one line must be written as
Testing your code:
Script will evaluate each code. I shall provide the input image (unseen) (size 20x20, gray
scale) to your code and will give +1 or 0 reward for each right and wrong prediction
respectively. Same test data set will be provided to each script.
10 freelancere byder i gennemsnit $156 på dette job
Hello! Your current dataset is very similar to MNIST dataset (28 x 28). We can use CNN to classify this type of handwritten letters. Python + Tensorflow will be used for this project! Thanks.
I have experience in CNN using python. Let me try it and then you can decide to hire me, if you are satisfied. I have also done a project on handwritten mathematical symbol recognition using python and Deep Learning.
I am shatrunjay Shukla, working with HCL technology ltd having total 11 years of experience. Key Skills: Asp.Net, Asp.Net MVC, Sql Server,css3, Web API,WCF,WCF REST, Web Service, Machine learning, DevOps,Python