CUB Lubdhok

https://www.youtube.com/watch?v=1bLFOUe8hQQ

https://plastic-crocodile-frontend.vercel.app

In our project named Plastic Crocodile, we have trained our ML model to monitor, detect, and quantify plastic marine debris. We have used Nasa image resources gathered from OR AND RS MARINE DEBRIS PROGRAM, we have separated those images into three categories of the testing dataset, the training dataset, and the validation dataset. We have manually labeled some of those images with plastic marine debris of different kinds. Using that training dataset, we have trained our Machine Learning model by adjusting and tuning the model hypermeters.

And finally, when the trained model is ready to detect the plastics marine debris, we have used the testing dataset as input images to detect the plastic marine debris object in a bounding box with the confidence level. Finally, we have uploaded those detected images with GPS location to our MongoDB database which we can be visualized through our web portal.

To develop these solutions, we have used Python, Machine Learning Yolov5 framework to train and detect the plastic marine debris object. And for the web portal, we have used React, Node-JS, and MongoDB databases to store and visualize the data.

Publicly hosted web portal link:

https://plastic-crocodile-frontend.vercel.app/

Frontend git repository:

https://github.com/solaimanshadin/plastic-crocodile-frontend 

Backend git repository:

https://github.com/solaimanshadin/plastic-crocodile-backend

We have made dataset using data from resources provide by NASA. We collected images from OR&R’s Marine Debris Program I NOAA and annotated the images in YOLO format to train our model.

Toxicological threats of plastic | EPA

MARINE DEBRIS FACTS I NOAA

Marine debris : Garbage patch Experiments | NOAA

FOR DATASET 

OR&R’s Marine Debris Program I NOAA

In an era where there is very little scope of the invention given to students, the NASA Space Apps Challenge allows us to experiment with an open heart and we get a chance to explore ourselves from within. The experience was just like going out of the books and revisiting the era of Einstein and Newton.

First of all, the journey was filled with excitement. A lot of things came out behind this work. We found out how teamwork can make a beautiful thing in a short time. We gathered experiences about how we can use our knowledge and how we can learn from others to complete any task. Most importantly this program or we can say the NASA Space App challenge gives us a platform where we can think to create something using NASA data that is unique.

Here we learn how to work together as a team. How to maintain everything step by step. Here we acquire knowledge regarding global climate issues and use these to solve environmental problems by using technology. So this platform gives us a real-time experience on how to use technology to solve environmental problems. Though it is teamwork sometimes we work virtually thus from this work we learned how to work virtually and complete a task on time. Most important thing is that we found a place where we can use our knowledge and gather knowledge from the members and we must say that the most valuable thing nowadays is data.

Our motive is to bring back a healthy marine environment, where there will be no plastic debris, no marine creatures will have to die because of plastic consumption. Hence, this inspired us to choose this project to protect the marine ecosystem from getting destroyed.

We have used Nasa image resources gathered from OR&R's MARINE DEBRIS PROGRAM, and we have separated those images into three categories: testing dataset, training dataset, and validation dataset. We have manually labeled some of those images with plastic marine debris of different kinds to prepare the training dataset. Using that training dataset, we have trained our Machine Learning model by adjusting and tuning the model hypermeters.

And finally, when the trained model is ready to detect the plastic marine debris, we have used the testing dataset as input images to detect the plastic marine debris object in a bounding box with the confidence level. Finally, we have uploaded those detected images with GPS location to our MongoDB database which we can visualize through our web portal.

First of all when we find any problems then we mark the problem or note it and discuss with team members how to solve it. After that, somehow we can manage the problem because teamwork can solve any problem easily. But there is a big problem that arrived when we needed a lot of data to get the accuracy. This problem was solved when we gathered data from NASA and processed our own dataset. We then combined our previous dataset with Trash Can 1.0. After that, we used the data that is needed to solve the problem. This dataset is accurate enough to detect marine plastic debris because now the accuracy of our app is high. It can detect marine debris and it can differentiate between marine animals and marine debris as well. We all are facing the Covid-19 pandemic so educational Institutes are closed in Bangladesh and we have to work from home so we solved it by working virtually because we have to work as a team.

We would like to thank NASA who gave us this kind of platform where we can participate every year and do something innovative and new. We would also like to thank our local host BASIS, who helped us a lot. They guided us on how to do this work, they also gave us a mentor. Our mentor guided us properly so that we can do our project that can make an impact. So our thanks go to our mentor. Our University teachers also helped us, guided us, and inspired us to participate. Mostly our Honorable faculties Faria Tabassum and Shah Reza Fahad. We would like to thank them as well. But big thanks again go to NASA. They created a big platform and they also gave us a lot of data so that we can use it as it is required.

For Data:

1. Toxicological threats of plastic | EPA
2. MARINE DEBRIS FACTS I NOAA
3. Marine debris : Garbage patch Experiments | NOAA
4. OR&R’s Marine Debris Program I NOAA
5. Trash Can Dataset 1.0
6. Own dataset: https://drive.google.com/folderview?id=176edjJK4V-wCY6Unc8cOb7026m70iBBA

To train ML models and detect plastic marine debris objects:

1. Python,
2. Machine Learning
3. Yolov5 framework

To store data and visualize the data:

1. React
2. Node-JS
3. MongoDB databases

#NASASpaceAppChallenge #CUBLubdhok #WaterPollution #Debris #PlasticDebris #MarineDebris #MarinePlasticDebris #GlobalWarming #PlasticDetection

This project has been submitted for consideration during the Judging process.