Getting onluy 44% accuracy in pokemon's challenge

#1

Here is my code - https://colab.research.google.com/drive/1KYIvs9mGi6Bn6hng0VgN_XnFvPW792LP?usp=sharing

#2

hey @sanchit123manchanda ,
your code is absolutely great but a bit long , try to optimize it.
and for the challenge , with SVM you don’t have much parameters , you can only optimize learning rate to improve your performance.
For this challenge , you can try different machine learning techniques available in sklearn package to perform better on this challenge , or more improvement can be attained from neural networks , more specifically CNN’s . There results on images are really amazing .

I hope this would have helped you in your doubt.
Thank You and Happy Learning :slightly_smiling_face:.

#3

I have tried many algorithms except deep learning and 44 % is the max I can get. Also when I tried the same algorithm on training data I got it around 63 %. and it remained exactly same when i did it with sklearn even after changing the parameters and trying different kernels.

#4

we have generally seen that conventional machine learning algorithms perform poor on images.
I would suggest you to go with deep learning techniques. It will surely give you better results.