Hi, I am new to deep learning. I am trying to make CNN on CSV data. The data is derived from the images. And I am predicting x,y,z coordinates for mono pong game. Now I am stuck at the preprocessing of the data. I have (667,225) CSV of input set and (667,3) CSV of labels. So I read through the internet but still struggling with reshaping the CSV. How should I convert CSV to tensor? The actual pixel of the image is (1920, 1440). (I do not have images.)
Can anyone tell me how can I prepare the dataset to feed it to the architecture?
and one more thing
In CNN architecture how we can decide the input node, hidden lawyer and kernel size?
I know there is lots of information on the internet but after reading it I became more confused.
Please help me out. Thanks in advance.
From your CSV shape, I would have to say your images might be 15x15, 667 is the number of images. So the shape of the tensor would be (667, 15, 15, 1) adhering to the format (batch_size, image_width, image_height, num_channel)
1 Like
Thank you so much for your reply. Labels do not need to change shape right?
There is no need to change the labels shapes.
1 Like