A Real-time Trash Detection Web App

1 in 4 items thrown in recycling bins isn’t recyclable, and this contamination costs the US recycling industry up to $500 million annually. I wanted to help solve this issue using Machine Learning to help the recycling industry.

Devpost:

Jupyter Notebook:

TIMESTAMPS:
0:00 Intro
1:16 Training Data
1:30 Object Detection Model
2:34 Model Training
3:32 Deploying Model with React

[Nimbugg]

Thank you, it is very helpfull for my project (Trash detection with car camera) it will help local authority/ municipal corporations to dispose it quickly and in efficient way .

[domstardc]

LOVE!! Should still be able to make this into a real world app

[bloxxer]

Great idea and great execution. Keep it up!

[srinivasgoud]

I am a btech student .. Me and my other 2 frends are doing the project on this 😮😮😮i thought the idea was unique but, it's already exist .. we thought of dropping this idea coz we don't know how to work on it 😢 after seeing your video i got some hope

[SamuelOrtizFishman]

Hey Alex that's a really cool app! I am looking to create something that uses ai trash detection for a personal project.

[bennguyen]

Any thoughts on how well AI can code such a task? For example, will future hackathon competitions disallow AI.. or maybe use AI, but it will be individuals (no teams)? Or is there some limit to how well AI can take an idea and do the design/front-end, and backend development?

I probably just don't know enough of all the different tool types (chatboxes, AI Agent frameworks, code assist) but personally, I've been trying to do an augmented-reality app (no glasses, just using the smartphone's camera) and have not found AI to be very useful.

[marcgentner]

I am currently building for school a same type of project. But with resnet50 as backbone. How did you vectorize your images and segmentations and bboxes?
I have scaled my images to a standard size together with the segmentations and bboxes. But my model cant seem to trian on the segmentations because the segmentation cordinates all have diffrent shapes. This means my resnet model does not accept the segmentation vectors. How can i tackle this problem or how do i vectorize the segmentation cordinates so my model knows how to work with this data?
Any help is welcome. Thanx in advance!

[redmonster_]

Honestly, i'm not sure how this is just at 200 views. You could add background music. Also, as Tom Scott once said, code is incredibly dull on camera. Keep it up !

[HungVu-uirh]

i really like this. But I don't know how to integrate taco into my folder. I unzipped it after downloading it and copied it to the project folder downloaded from GitHub but when I run the notebook jupyter, at the line "taco = pd.read_csv('taco\\meta_df.csv')\n", "taco.head() there is no meta_df. Please help me...

[odd-rob]

Is there any way to get the model and code from the project to test this out myself cause this is awesome!

[Masxaiir]

whan is the next dance tutorial video??