How can we achieve 90+ accuracy?

#1

I have seen that many people in the leaderboard have achieved 96 , 100% accuracy etc.

Is there a solution for this problem in github or somewhere…I have completed my challenge and I got 77% maximum after several tries.

#2

hey @lintasheelkumar ,
Firstly, the ones who have achieved score as 100 are actually our mentors and testers only , while testing the platform they had submitted the original files and hence got 100 score.

For others , we don’t have any such solution or code which we can share to get higher score.
Some tips to improve score :

  1. try various algorithms , mostly while we are working on images , neural networks with CNN layers will work much better.
  2. Try transfer learning ( using pre trained models. )
  3. Try models ensembling .
  4. As the data set is not that big, so try using data augmentation, it would help the model learn images much better.

I hope this helps you and you much better score.
Thank You and Happy Learning :slightly_smiling_face:.

#3

Ok Thanks…What pre tarined model is best for these problems? Is there some criteria in choosing these models?

#4

Basically it depends upon the mode architecture ,
for example if your data contains an in-depth details then using small networks like vgg or alexnet wont be much useful. Hence then we use bigger models , like densenet or Resnet101 , they can extract very deep features of the data and hence at that time they can be used.

#5

I hope I’ve cleared your doubt. I ask you to please rate your experience here
Your feedback is very important. It helps us improve our platform and hence provide you
the learning experience you deserve.

On the off chance, you still have some questions or not find the answers satisfactory, you may reopen
the doubt.