The project has created by our team aims to improve the estimation of vegetation index value collected by NASA Settialites and utilize them for benefit of farmers and in the long run benefit of humanity to feed starving people or utilizing agriculture with collected data.

What exactly does it do?

The model created by STRONK team estimates the vegetation index with collected satellite data and gives min value, max value, mean value, median value, std value, p10, p90, and cloud coverage for ndvi values as output and with those values creating a fine gaussian curve of probability is possible to perform statistical test like A/B testing.

H0 : vegies are same

H1 : vegies are different

H1.1: vegies will grow less

H1.2: vegies will grow better

How does it work?

The model works like every other deep learning model or ml model for x input gives y output.

What benefits does it have?

The estimation of the model has been heightened by newly be created AI architecture called MFP by our team. MFP utilizes KAN that newly created neural network algorithm. Someone who has experience with KAN will understand KAN's problems but, MFP solves those problems like the high number of parameters, ability to low function at wrong hyperparameters, etc. Therefore, our AI newly invented AI architecture will outperform classical linear layered models and KAN. Also, we invented a new Transformer Architecture but that will be overkill for a demo model so, we did not add it, and all of them will be published as papers at conferences or journals.

But the most important aspect of that model is interpretability with simple algebraic logic equations which can be provable by math.

other benefits:

1. high functioning new architecture for dl, nlp, ml, drl

2. high estimation power for vegetation index for next years that can be fed with data that spread over more than 30 years.

3. compare ml imputing algorithms on continuous variables

What do you hope to achieve?

1. STRONK team wants to achieve an estimation of the vegetation index to calculate vegetation probability of growth like the Gaussian Curve. Because, the model estimates min value, max value, mean value, median value, std value, p10, p90, and cloud coverage so, farmers will have a fine curve of probability for estimation.

2. Showing performances of ml nan value imputation algorithms

3. Announce MFP, new Transformer, and other Deep Learning Architectures that are created or invented by the STRONK team will be beneficial for humanity.

Tools

1. Python

2. Pytorch

3. torchmetrics

4. torchinfo

5. numpy

6. pandas

7. sklearn

8. DALL-E

9. EO Browser

10. Climate Data Store

11. Tensorboard

2024 NASA Space Apps Challenge

STRONK