th small regret as can be made out from my subject
My best public score got my rank of 90 ,se ond best unofficial 50th rank
.I built solution using newest efficient b5 and b5 model built by Google
Hope it can be incorporated soon in fast ai
.thanks Jeremy and sylvian for helping deep learning aspirant built top 5/ 10 solution for kaggle competiton .
I still remember Jeremy words during initial lectures that people who are ranked in top 10 of kaggle competition know what they doing and why they doing:)…
Congrats ! Personally I managed to got in the Top 77%, so probably I have no idea what I am doing. But I am still at the lesson 3 of the Part1, so this competition was my homework of lesson 1 and 2 x) I just regret to not have spend more time on it (i.e. start the lesson before :p).
I can wait to see what have done the first and second position. My bet is that they used an efficientnet and a really qell stratified validation set.