What can be the possible reasons for validation accuracy remaining constant(0.5)? I have approximately 207 normal images and 179 covid images in training, and 25 each in test. I have rescaled the input images by 255 as well. and the model was fit using data generator.
Validation accuracy remains constant
hey @ambika11,
There are many reasons for validation accuracy to not change ,
- Your model is getting overfitted, it is learning the data in such a way that it is not able to work on validation well.
- Lower Learning rate , the model is isn’t learning much to improve performance.
- Model has reached convergence , or might have been stuck in local minima.
- Model has been regularized a lot , due to this , it is not able improve itself.
To get the proper understanding , you need to have a look at the history plots of model .
How does its loss , accuracy validation loss and validation accuracy changes with time , helps to understand the model performance.
I hope this would have helped you understand it.
Thank You and Happy Learning 
.