Recycle robot – using raspberry pi and tensorflow | Hacker News Reader

Recycle robot – using raspberry pi and tensorflow

recycleai.tumblr.com

167 pointsgerrypez12335 comments

My main problem with recycling as done today in most cities I just don't know what do they expect from me. In London my district had different bins for different colors of glass. I live in NYC now, where I hear they have a 2 stream system, one for metal and glass, one for paper and plastic but most public buildings just have "recycle" and "trash". Should I throw a plastic bottle in there? I know you can recycle some plastics but not all. Cardboard is obviously recycled, but can I put in other paper boxes (not necessarily corrugated) between the cardboard? What about my paper coffee cup? Is that compostable or recyclable paper or trash?

In a weird way a machine like this would solve my issue if it were programmed by the building or city to do the right thing...

Edit: typos

Here the rule for paper is clean paper (no wax, no food residue, etc) and only corrugated cardboard (no boxboard). That's based on how useful the materials are so I guess is a pretty common standard.

I have friends in Europe, and last time I stayed with them they tried to explain, as best they understood it, the local recycling scheme. It was so complex that packaging has something like 5 different recycle logos, and their municipality only accepts 3 of them, but they're all just a circle of arrows pointing at each other with a number in the middle.

So can you recycle the brown bottle with the more square shaped arrows and the "3" in the middle? How about the next bottle exactly the same with the "3" in the middle but slightly more swirly arrows? Or the same square arrows and a "4"? They didn't even know, so they presort, show up at the recycle center with everything and get rid of what the employees there will accept and take home the rest for regular trash.

It's a huge waste of everybody's time. And the 20 minute drive to the local recycle center can't be helpful.

Most places in Europe I have lived will pick up everything from public containers not further than a few tens of meters from your house.

You only need to drive to recycle very-specific trash like baby diapers, dangerous substances, furniture and big appliances.

Common division is as follows:

- Blue for paper and carton - Green for glass (some places classify glass in three colors: green, brown and white) - Brown for compost - Yellow for Metal and Plastics - Other: Anything that does not fall on the previous containers.

Exceptions: batteries, electronics and dangerous substances, which, as I explained before, need to be picked or brought to a special recycle center.

Are you talking about plastic and their resin codes? Only the number there matters (it identifies the type of plastic), the arrows not. https://en.wikipedia.org/wiki/Resin_identification_code

I am yet to be convinced that domestic recycling is the way to go, as opposed to putting all the trash to one huge landfill and subsequently unleashing mining robots on it.

It seems to me that sorting stuff (recycling) at home is prone to errors to such a high degree that it is just a feel-good thing, as opposed to something useful for the environment.

In Cambridge, UK, we get somewhere close to that with 3 bins:

- Green, for compostable waste (food, leaves, shredded paper, etc.)

- Blue, for recyclable waste (pretty much anything made of plastic, foil, paper/cardboard or glass)

- Black, for everything else

The only things you really have to watch out for are things like plasticised foil and styrofoam. Batteries can be tied to the top of the blue bin in a plastic bag.

It works really well, and you basically just have to know what's recyclable and what isn't (which is usually written on the packaging anyway).

I think our local authority is a bit ahead of the rest of the country on this, though. Most places in the UK still make you sort glass from paper, etc.

This is my local system too, and it seems eminently sensible. Worth noting that the blue stuff will go through a process similar to the Raspberry Pi sorter, but on an industrial scale, so it's somewhat similar to the idea of sending mining robots into separate things out after the fact, but made much easier by not having food and babies nappies smeared all over the cardboard boxes.

In Vienna, Austria there is an incineration plant (https://en.wikipedia.org/wiki/Incineration) which burns trash to produce energy. As I understand it, they can't just burn everything at once otherwise they can't control the combustion temperature, which then produces harmful byproducts which must be cleaned out between burns, thus reducing the effectiveness. Perhaps a few plastic bottles here and there mixed inside isn't so bad, but if everyone were to do it, it would cause issues? I do know there are "trash-checkers" who dig through trash cans to find, document and fine offenders (it's apparently a fairly well-paid job too).

I think that an expectation of global standards is unrealistic. When you arrive in a new city, you have to learn plenty of things like where things are, how to get around, and what laws differ. I don't think that asking people to figure out what should go in which bin in the city you just moved to (an event that typically occurs annually at most and a question that takes a few minutes of Googling to answer) is excessive; it's just part of the new city learning curve.

Awesome! Kind of like a microscopic version of a full scaled recycling sorter facility.. https://www.youtube.com/watch?v=SIVKmwzWSuc

That video was fascinating! Thanks for posting it, I've wondered how those facilities work.

Very cool - it's really interesting to see the speed difference between OPs project and the industrial version (thousands of pieces a minute).

Take a look at https://youtu.be/SIVKmwzWSuc?t=73

for the "first optical sorter".

Wow! This is a fantastic application of Tensorflow. I am a mechanical engineer and I am constantly thinking of ways I can use AI/ML in my systems/work. I think that there is tremendous opportunity for using tools like Tensorflow in unsexy applications such as recycling.

If the maker is around: Open Food Facts has an insanely large db with recycling info, barcodes as opendata.

pierre at openfoodfacts dot org

I'm not the maker, but that's super helpful Pierre!

One idea could be, use this dataset for training an image recognition net, and when there is no codebar fall back on that.

I was thinking on my morning walk about making a small drone that could fly through an area looking for trash it can pick up.

It could potentially go places that are difficult for humans to get to pick up some litter, like trash in streams or thickets.

Drone Wall-E?

I'd love to see this replace the current multi-bin setups I see most places. The ones where you spend so much time trying to figure out which is the right bin that inevitably some people just dump everything in the easiest location. It doesn't take many mistakes to ruin the efficacy.

An automatic sorter could solve that and training could be tailored to the location.

Starbucks seems like a good pilot location.

This problem is solved with off site recycling. People get it wrong either through ignorance or laziness. Just dump it all in one bin and sort it at the facility. It's cheaper and more accurate. Add an arm to pick things off a conveyor belt, then you'd have something.

Yes, but ...

It provides some jobs, yes. But mostly low-wage.

Broken glass is a key issue. It's especially hard to separate from paper. I vaguely recall that silicon is also an issue for aluminum.

If your skills are limited binocular vision, object recognition, and an opposable thumb and n 2017, you're screwed. Hell, you've been screwed for 30 years. Everyone knows this.

Yes, and that's the problem. The US had welcomed them, because lots of dumb work needed doing. But now, not so much.

So how do societies deal with them? People with no prospects, especially young men, are dangerous.

Government make-work programs, but they're pretty much boondoggles by design, because you're intentionally not optimizing for cost-benefit? But rather maximizing headcount.

If I was in charge retrain people under 50 for something, with more education subsidies (including free post-secondary education) increasing as Age decreases (eg an18 year old goes to college, a 45 year old learns welding) but with the caveat that you have to move out of Methlandia. (Relocation package provided.)[0] People over 50 or those that don't take the offer get free carfentanil.

[0] https://newrepublic.com/article/131743/poor-get-trapped-depr...

I'm new to AI/ML, so it would be interesting to see the source code used for this.

Not bad at all.

A little self promotion: I'm part of a startup making robotics for recycling... maybe a little further along than this project :-) but it's a great start!

https://resource-recycling.com/recycling/2017/03/21/carton-p...

http://amprobotics.com/

We don't need a robot to do it for us, we need an environment that teaches us to do ourselves because we should. A better arrangement would be for the pi to light a simple led lamp over the bin it is supposed to go in, and let the human learn to do it themselves from habit. Otherwise it is an impressive engineering project you put forth.

Really cool project, this inspired me!

FYI: ZenRobotics has been developing waste separation technology for years now. Not sure on what stage it is now but according to their website, they seem to have some working solutions already. http://zenrobotics.com/

This is something of great interest to me, though I don't have the wherewithal to pull something off like this. I'm glad someone did.

Although this is at a hobbyist scale, I think it might develop into micro-recycling sorting, one of the pieces to help increase the rate in which we recycle goods.

Neat project! Love it.

Sort of sad to see the most useful object to recycle (the steel scissors) went to the landfill. The recycle value of that one item would have dwarfed the value of a year's worth of banana peels.

Impressive, it's already more intelligent than half the people in my apartment building.

What a cool project. Nice!

Love to see code...

Reply on news.ycombinator.com