Lacuna - Correct Field Detection Challenge
$10,000 USD
Can you design a method to accurately find field locations?
259 data scientists enrolled, 32 on the leaderboard
AgriculturePredictionSatellite
26 March—13 June
Ends in 2 months
Mapping Constants
published 30 Mar 2021, 13:40

Hi Zindi,

In the starter notebook, there are mapping Constants to get the corresponding pixels for Planet images (84 by 84). How were these constants derived?

And what would those constants be if one needs to map onto Sentinel-2 (41 by 41) instead?

Kindly help clarify this.

Hi Alchemi,

The constants are the normalized scaling factors of the Longitude & Latitude of the original bounding box on the map.

Therefore they can be used for both the RGB and the Sentinel images, while adjusting the image size accordingly.

In the starter notebook the conversion from raw (x,y) to coordinates for the RBG is:

x1 = x0 - np.round(row.x.values[0] / CONST_X * img_jun17.shape[0])

y1 = y0 + np.round(row.y.values[0] / CONST_Y * img_jun17.shape[1])

Where img_jun17 size is usually 84 * 84.

For the Sentinel-2, the conversion would be:

x1 = x0 - np.round(row.x.values[0] / CONST_X * img_sentinel.shape[0])

y1 = y0 + np.round(row.y.values[0] / CONST_Y * img_sentinel.shape[1])

Where img_sentinel size is usually 41 * 41, using the same CONST_X & CONST_Y.

I hope this answers your question.

Yes, it does. Thank you very much