- Final model gives 94% accuracy in classificaiton of dogs and cats dataset
- This dataset is taken from kaggle
- I have trained the model by downloading the entire dataset, however trained with 1000 cats and 1000 dogs due to unsufficient computational resources
- converted all the Images to square 200 x 200 images
- I have used
one block VGG
,two block VGG
,three block VGG
, plotted the graphs betweenloss,val_loss
,accuracy,val_accuracy
and figured out, at which epoch , the overfitting occurs
- Data Augmentation is one of the regression technique, and it has improved the accuracy to around 60%.
- Dropout is also a regularization technique, however it gave a accuracy of only 48%
- Final model used is VGG 16, using transfer learning , I have trained the model and achieved a accuracy of 96% in classification of dogs and cats .
- Using Sample Image, I have tested the model and the Final Model predicted correctly