Future improvements and possibilities

This project is still a Work In Progress.

Improve accuracy

As can be seen from the dashboard snapshots, the crop area coverage is very inconsistent. This occurs due to the fact that the analysis by the computer vision model is not accurate.

This is because the IBM-Nasa model was trained primarily on satellite images of US regions and are very bad at analyzing images of other regions.

The satellite images also have to be properly scaled and normalized for the model to be able to analyze them.

All this can be corrected by developing our own computer vision model and/or training the model on AMEA region data.

Integration with Field Segmentation

Field Segmentation is a project done by the intern team showing the segmentation of regions into component areas which can be treated individually, based on sattelite images of the region.

Both these projects can be integrated into one.

The field can first be segmented into individual regions and then crop area coverage can be detected for these individual regions.

Or, crops can be identified for an area and then this area can be segmented into individual regions based on crop spread.