aryangoyal7 / deep-learning-hello-foss Goto Github PK

You might have noticed that our accuracy didn't saturate after a certain number of iterations, try increasing the number of epochs to see if just doing that gives more accuracy.

We've noticed the mean and STD values may not be correct for normalizing the images, hence calculate mean and standard - deviation for each channel and fill in the correct values.

Try to realize how size in batch loader size effects the performance of the model for a specific dataset, and find the ideal size for our model and data to optimize the accuracy.

In the model data splitting for validation seems off, 50 percent for training and 50 percent for testing, fix it for better training

it's easier to see what's happening in our model when we visualize the results, plot various graphs like accuracy vs number of epochs, loss vs no, of epochs,etc or check with individual images to see how well the model is doing