Agriculture Crop Segmentation

Applying Image segmentation on Satellite images

Around two years back, I worked on a small but interesting use case. I was given two .npy file and asked to predict the type of crops in the image. This was the first time I saw this kind of use case and it definitely felt interesting. I was not sure If I can solve this, and in three days that I spent on this problem, at least three time I felt it can’t be solved but at last I got this output:

I felt good actually seeing this. Our model filled the missing (black) space and the interesting part was that along with metrics, we can visually see the prediction performance and computer vision as a whole is interesting field from this perspective because seeing good numbers in metrics is one thing and visually seeing the predictions is another thing. Our eyes is a tool trained for millions of years and as an analyst, I think there is no harm to get some extra shot of dopamine as long as we are backed with numbers.

Please find the code here: https://nbviewer.org/gist/sparsh-ai/55e7af650b740c03f940844e1911fb7d

Enjoy!