hey i’m getting very low accuracy 10% in this challenge.I’m using transfer learning for this.I’m has earlier also messaged 3 days ago but there was no response since then.Pls help me to improve accuracy.suggest me som echanges in ntbk.Link to ntbk-
https://colab.research.google.com/drive/1PFqBjr1WUquKHwp_A4gy_KkyDs-bzBny
PokeDox Challenge
Hey @Nik1, another TA has acknowldged your doubt. May be he forgot to reply. Dataset has been updated from the problem set very recently, I will check what’s the accuracy achievable on the challenge. And than i will check your ipynb as well and will get back to you very soon.
By the time have patience, since i have to train a model myself, and it will take a while.
Hey @Nik1, you have to make predictions in the order as in sample_submission.csv and not in any random order. Plz check if you have made predictions in right order.
Also you have used your own label dictionary, use provided in the problem statement,
0 : “Aerodactyl”,
1 : “Bulbasaur”,
2 : “Charmander”,
3 : “Dratini”,
4 : “Fearow”,
5 : “Meowth”,
6 : “Pikachu”,
7 : “Psyduck”,
8 : “Spearow”,
9 : “Squirtle”
Also normalize images by /255.0.
You will be able to achieve good accuracy by all these changes.
Hope this resolved your doubt.
Plz mark the doubt as resolved in my doubts section. 