Neural Network Challenge-: House price prediction

The dataset is available here in prateek bhaiya repo. Folder name is dataset and in it subfolder named as images.

hey @sambher_shubham ,
why are you limiting training steps to 7 and validation to 4 ?

Because i have more number of images in training and less in validation

Thats not any problem ,
specifying them means , you only need to train on 7 batches i.e 7*32 = 224 images in on epoch , whereas you have 512 images for training.
and same for validation also.

ok i get the point. I am changing the batch _size and steps for the training part


My validation accuracy is only 78% but prateek bhaiya accuracy is 97 in detecting covid-19 using X-rays. The only change that i have made is that the training and validation images has been increased for both normal and the covid xray images. But despite that my accuracy is 78%. Can you plzz tell why this is happening?

I have increases the number of epochs also but still it lies between 77 to 85

While predicting the model is predicting the correct class for the covid xrays but for the normal onces it is giving me wrong class in most of the examples.

hey @sambher_shubham ,
can you please provide me with the link to this dataset , as i need to try it once to understand what actually the problem is.

Link of covid dataset

Link of normal xray dataset

I have extrated the covid 19 images from the covid dataset you can see in my code

Have you seen bhaiya?

hey @sambher_shubham ,
sorry to be so late , due to some work , i wasn’t able to check it.
Working on it now. Will reply you as soon as i get something.

hey please let me know which images are there in covid_dataset/Train folder path.

and don’t initialize steps per epoch while you use generators.
You are not training your model on whole dataset. kindly remove that line and validation steps too. and then check how the model performs.

In the train dataset folder there is folder named covid with covid images and there is a folder named normal containing normal images

I have moved steps per epoch and validation steps also. I have trained my model for 20 epochs and it is still giving me 82 percent accuracy

The accuracy is fluctuating in each epoch sometimes goes around 88 sometimes between 70 to 80