CGIAR Wheat Growth Stage Challenge by CGIAR Platform for Big Data in Agriculture
$3,000 USD
Predict which phase of growth wheat crops are in using photos taken by farmers in India
459 data scientists enrolled, 204 on the leaderboard
AgricultureComputer VisionPredictionImageSDG2
28 August—4 October
38 days
How I obtained a Leaderboard Score of 0.44
published 3 Oct 2020, 23:54

Firstly, I would like to thank my mentor Karim Amer@karimAmer and @Johnowhitaker for their generous and kind attitude in letting others learn.

Am actually a newbie to computer vision, I started less than Four weeks ago, but by garnering knowledge from @KarimAmer and @Johnowhitaker, I was able to obtain a leaderboard score of 0.44.

You can find the code here

Since am a newbie, there might be some bugs in my code. If you find any, do let me know. Also , I am open to ways of improving.

Congratuations to the forthcoming Winners.

Why do you share your solution now? This competition has only one day left. You should not share your solution at the closing moments of a competition.

it does not matter. I am sharing so everyone can make good use of it while it still lasts. Am sure posting my solution on the disccussion forum isn't against ZINDI rules or am I wrong?

I don't think it is against any rules that I am aware of. In fact sharing solution is good so that everyone can benefit from it. But not at the closing moments. You are just giving away freeloaders a chance to be at the leaderboard which will be injustice to the participants around score 0.44 who worked hard for the last couple of days. This is just my opinion , I don't mean to offend you in any ways or to discourage you from sharing your solution.

Well you are right in a way, what I was hoping for was people who were working and were stuck could see means of improving, probably get to the first five, also I wasn't hoping to share it this late. Am sorry if it happened few hours to the end of the competition.

you share far before or after; not in dieing moment

Thanks for sharing. It's really helpful!

thank you @Zion_ for sharing your work. I am also a newbie to deeplearning. I had a doubt on selecting the training images.

Did you take both good and bad quality images to train your model ? . When I use both quality images for training my resnet50 network using tensorflow. The loss for the test set resulted in 1.96 (approx). Can you give any idea to improve my model and it is very helpful for my learning process.