Medium blogpost about kaggle fisheries competition

Hi, I’ve wrote a blgopost about my experiences from participating in kaggle fisheries competition.

any kind of feedback will be welcomed :slight_smile:

1 Like

hey @shgidi,

thanks for sharing your experience. Your blog post sounds like that the challenge was quite harder than expected due to the strange public leadership sample.

Congratulation on the place/challenge and your blogpost.

A short feedback (my opinion):
I like a small summary in the beginning. You mention you learned some lessons, but summarize them quickly. I think blogposts is more about the reader than the writer - why should I read the full blogpost? What is the key message / the most important points, the reader should keep? Normally people do it in the end, but I think it is a good start, so I know what is it about. In the academic papers, you read the abstract before and then decide if you read the full paper.
Example could me “methods”, the lessons you’ve learned, the unprecentable leadershipboard, etc.

Do you have more information, why the cropping out + use VGG on the cropped out image didn’t work? It was mentioned in the lecture, that it could improve the results.


Hi Benedikt, thank you for your feedback!

the cropping technique did work. the two high ranked solutions which i’ve linked in the bottom of the post actually used very similar architecture. However, in the course of training, me and my partner were discouraged from our results, since they were not getting us ahead in the leaderboard. I believe that if we’d dedicate some more work to fine tune the classifier, and tackle some detector issues, such as what to do if there is no fish, we would finish well above the sample-submission (30th place).