We think that AI will be able to identify some common and obvious types of fruit, but not different varieties of the same fruit.

The fruit consumption in Canada per year in 2022 is 5,000,000 average apples, which means,  33,000,000,000 stickers, that end up going into landfills and collecting.

Vinyl takes up to 1000 years to decompose, like plastic.

Plastic eventually breaks down and keeps breaking down until they are micro sized hence the name micro plastics. These get in the water and ground, and affect animals because they could mistakenly eat it for food, accidentally swallow it, or ingest it by trophic transfer, meaning that if another animal eats that animal, they could ingest the plastics too. Animals could get very sick and die from it.

In our home compost, fruit stickers tossed away by the previous owner were still clearly visible and intact from 20 years ago. The ink and words were still readable.

Compostable sticker designs are twice as much as plastic stickers

You need the world to adapt to change. If only one company makes compostable stickers, you need the whole world to make a difference.  Because of low demand, it is likely that these compostable stickers will only be popular once people recognize it. This could take a while, since people haven’t realized, or are ignoring the harm it does.

Compostable stickers are not yet widely recognized as an unnecessary single-use plastic. There is a “lack of demand and production”.

In Washington, facilities identified non-compostable produce stickers as one of the top five things that come up.

Composting has lots of benefits for the environment, including helping to divert food waste out of landfill, where they make methane. Composting helps other things grow, because of the nutrients that were once natural things that got tossed in the soil.

By addressing one of composters’ biggest contaminants, the compostable packaging and produce industries would be supporting the profitability and viability of the composting industry. This would help create a more positive relationship between groups, demonstrating care for the quality of finished compost, and serve as an example for other compostable products. 

In 2017, a Washington State Organics Contamination Reduction Workgroup report determined that “although compostable stickers were made and working, it has not yet reached a demand to take over plastic stickers”. Since that time, the landscape of manufacturers has not changed their ways, although more retailers have trialled solutions. Presumably, this means that today’s compostable produce stickers are also cost-prohibitive. As more solutions come on the market and are adopted more widely, they become more cost-effective to manufacture. 

Bans on plastic produce stickers are set to go into effect, consumers are likely to grow increasingly frustrated, and composters may become stricter about not accepting stickered food waste.

• Compostable products will break down within the time needed by the composting environment and will not release harmful residues.
• Biodegradable products have no such definition, and the term is not regulated.

Eco friendly terms can also be used to imply that a product is good for the environment when it is not.

Plastic Stickers:

Used on a variety of produce, the stickers are helpful at checkout because they carry important information: price look-up codes, or PLU codes. The International Federation for Produce Standards determines these PLU codes, which have been in use since 1990, with 1400+ codes assigned.

They cannot be removed with the equipment that is often used to remove other types of contaminants, such as Trommel screens or depackaging equipment

Produce stickers are also a very common contaminant in the residential food waste stream, since consumers may not remember to remove them from peels and skins, or they didn’t bother.

In the case of grocery store food waste, plastic produce stickers make it challenging to accept and process large quantities of off-spec or spoiled produce. This can result in truckloads of produce being turned away from composting facilities and instead sent to landfills.

Countries are attempting to change this. France became the first country to put a law in effect. They will not allow any more plastic produce stickers. This law will go into effect on January 1, 2022. In the United States, a proposed ban on non-compostable produce stickers is included in the proposed Break Free From Plastic Pollution Act In New Zealand, the Auckland City Council has also proposed a ban on non-compostable produce stickers.
 

Scott Amron the inventor of the ‘Brush and Rinse Toothbrush’ invented the dissolving soap fruit stickers. The project was called ‘Fruit Wash’. He wanted something that could still work as a barcode but have some other purpose that is okay for the environment. The stickers dissolve into a soap to clean fruit or vegetables from wax, pesticides or dirt, when in running water. Scott Amron made them so they still stick. It is made of ‘natural or organic’ ingredients. (Won’t label what it is made of). 

• Food experts say that the fruit soap stickers don’t clean much better than just plain water

In 2007, researchers at the Institute of Agricultural and Environmental Sciences at Tennessee State University tested the stickers on both diluted vinegar and plain water. Sandria Godwin, who oversaw the project says that the stickers were not necessary nor effective. After the experiment, Sandria Godwin says that she still thinks that scrubbing with water is the best way to clean fruit, but she likes the idea of ‘no waste’ stickers.

• The US Centers for Disease Control and Prevention also agrees that washing the fruits with water is best.

The stickers also tell the PLU or ‘price look up’ on a fruit. Because of that, it makes the inventory check-out easier, speeds up the process and the cashier doesn’t have to remember stuff.

• Don’t eat the stickers. This led to someone eating it by accident and having a really long cough and lung infection, until they removed it after two years.
• The stickers won’t break down in a composting facility. The fruit would essentially ‘melt’ in the heat, but the stickers would not break down because of plastic. The stickers can make it through the whole process of composting things (shredder, heat, and filter)
• The filter, which was ⅛ of an inch, didn’t even catch any stickers.
•  

Structures and Materials:

The fruit stickers are made of vinyl plastic, though sometimes paper is used. This is so that the stickers are water-resistant, and are not damaged by packing and transit, but they are also not compostable or recyclable.

The glue is non-toxic and edible, and leaves no residue on the fruit.

Other Designs:

Laser Imprinting:

Add both letters and images to a piece of fruit or vegetable by removing the pigment from the peel’s outer layer

 It is a superficial process that does not affect the product’s flavor, aroma, or shelf life, and the laser-tagged part remains edible. 

According to the EU-funded project, laser labeling could actually be faster than the standard method.

In 2016, Netherlands-based laser technology company Eosta and Swedish supermarket ICA ran a trial of organic fruit and vegetables with natural branding, which they claimed replaced millions of pieces of plastic packaging

Around the world countries are making the switch.

The Oregon State University Food Innovation Center developed a laser imprinting system that imprints a code on the fruit. They burn a print on the skin of a fruit. They work on a wide range of produce.

• These laser prints avoid the use of glue, plastic, and inks on fruits. 
• The lasers etch the brand, variety, and a code, similar to a QR code but different than the normal barcode, on the fruit. These are about the same size as an actual fruit sticker.

Process:

The produce goes on a conveyor belt and in the middle there is a sensor/scanner. It scans the fruit and then burns a mark on the fruit. The fruit is not damaged and the shelf life is the same.

Cons: 

It requires a huge change that most people cannot afford nor think that it is quite as efficient. Laser printing is not suitable for delicate produce. Some of the laser printing technologies may not work on oranges, grapefruit, lemons, etc, because of their skin, which may recover from the burn, by healing and covering the burn. It is possible to fix this by spraying liquid on the fruit skin after the mark has been made by the laser to trigger a reaction that becomes visible. 

Ink-Printing:

Printing ink on the fruit or vegetable directly, avoids the cost of materials and adhesives. Vegetable ink-based tattoos are considered safe and do not pose any health concerns. Capexo in France, has developed a process for printing food-grade ink, that works for most fruit. It works well for fruits and vegetables with a relatively smooth skin, such as mangoes. This technology would not work as well for other types of more rough Since technology is advancing with robots and AI, we feel that using AI to identify fruits in our experiment would fit into the AI dominance. It was something that we had readily available to us, and certain platforms were free to the public. AI is already in lots of places that you may not think of, including healthcare, transportation, education, and Google Maps, and does many things. For example, in healthcare, AI analyzes the patient's medical history, and tells whether or not the patient is healthy, and if there is any risks., such as pineapple or avocado. They are not water-proof, and could potentially rub off, when in contact with water.
Compostable Stickers:

Compostable produce stickers are not yet widely available. However, at least two manufacturers have market-ready compostable produce stickers. In 2015, Elevate Packaging, a sustainable packaging and label technology company, entered into an exclusive distribution agreement for North America with Bio4Life, a Dutch company that produces compostable adhesive products and labels under the brand name PURE Labels

Sinclair, a large produce labeling company, has been trialing compostable stickers since 2008, and produced its first 100% certified compostable sticker, called EcoLabel, in 2019

Compostable stickers design:

PURE Labels compostable stickers are both biodegradable and compostable

Their performance is almost exactly the same as their plastic counterparts

Their products are made of cello and are ASTM D6400 compliant.

Types of adhesives

• Pressure Sensitive Adhesive

Pressure sensitive adhesive. This adhesive bonds/sticks when pressure is applied.

Ex. Stickers, household tape, bandaids.

• Hot Melt Adhesive

This type sticks when moisture or liquid is removed. For example when water evaporates. Ex. School Glue.

• Drying Adhesive

Usually a plastic, which is melted and then cooled, which forms a hard, strong bond. 

Ex. Hot glue gun.

• Reactive Adhesive

When a substance or chemical is applied to this, it forms a bond. Most commonly is water activated adhesives. Water is applied, and the material is usually a plant starch.

Ex.  An envelope when you lick it and seal it.

Compostable Adhesives:

Some are made with animal proteins, or plant proteins.

AI:

Artificial Intelligence is a collection of technologies that allow computers to perform tasks that normally need human intelligence to be done, such as learning, reasoning, problem solving, and interacting with others. AI uses algorithms, data, and more to simulate human intelligence, allowing them to do their job. AI systems can learn from experience and improve their performance, a critical part of their programming. Data is crucial in the programming and learning process of AI, since that is how they get all their knowledge and information they need to execute their task and serve their purpose. 

Artificial Intelligence image recognition:

Using machine learning to analyze images that we feed it. It compares it to data and previous images that they have successfully identified to place it. The longer AI has been trained, the more accurate it is, just like humans. It can identify humans, objects, places, actions, and more, but might be inaccurate if they were not trained to recognize the thing you are showing it.

 

Google Gemini is a chatbot. It helps with advice, research, image recognition, and can come up with stories and images.

Logmeal is a platform that is meant to identify foods of all types. It is not meant to identify varieties of fruits. 

 

In order to have the results as accurate as possible, we have made our set-up as similar to a checkout machine, and its surroundings as possible. We wanted to have a good lighting source and a simple background, as well as a metal surface. We put the back of our old trifold, which would act as that white background. A metal baking pan would act as the metal weight scanner, and we tried to have the lighting not too dark. We used a phone to act as the camera scanner, with a stand so it stays in position for each photo. It mimics realistic grocery setting. A check-out scanner with scale, and fruit in plastic grocery bags

Controlled variables: 

• Lighting 
• Table position
• Baking pan position
• Camera position, 
• Trifold position
• Artificial Intelligence platforms (Google Gemini, Logmeal)

Manipulated variables: 

• V1- Fruit Type
• V2- Fruit Variety
• V3- Bag types (Clear bag vs Translucent bag vs No bag)
• V4- Angle
• V5- AI Softwares.

Responding variables: 

• The results from the AI softwares.

1. Purchase and gather any fruits or materials you may need for this experiment.
2. Pick an area or room, which has sufficient light.
3. Get a sturdy, flat table that is large enough to place your baking pan on.
4. Place your table against the wall.
5. Get your trifold and place the backside between the table and wall. Make sure your trifold is large enough to cover the whole shot. This will be a good background and not distracting.
6. Place your metal baking pan in the center of the table, upside down, in order to create a flat surface.
7. Attach the phone holder on the left side of the table, near the edge.
8. Bend the phone holder, so it is pointing directly at the set-up, horizontally.
9. Place the phone in the holder and make sure it is secure and won’t move around.
10. Get your plastic bags ready to use.
11. Get your first fruit and place it in the middle, and take a picture of whatever variables you want, making sure you keep the phone steady and in the same position for each shot. (bags, angle, type, etc.)
12. Keep the set-up the same, while changing out the fruits and variables.
13. Once you have all of your photo variables, input each individual picture file into Google Gemini and Logmeal, so that there will be two opinions about the picture for each set of pictures.
14. Gather all your information into a chart or table.

Variable 1- Fruit Types:

Google Gemini successfully identified all the fruits correctly 7/7, though for the Navel orange, it only put “orange”, while Logmeal, only identified 5/7 correctly. Logmeal mixed up peaches with nectarines and we think that is because a peach and nectarine are similar, but a peach is fuzzier. We think that the pictures may not have been clear enough and the Logmeal could not see the fuzziness. Logmeal also could not identify a nectarine and responded by saying it was an apple. We think that because nectarines and apples are both fuzzy, Logmeal mixed them up, since you could not see how big they are in the picture unless put side by side. Logmeal was not very specific about the varieties of fruits.

Percentage of correct pictures: Gemini - 100% Logmeal - 71.43%

(% calculation: # of correct pictures / total pictures)

 

Variable 2- Fruit Variety:

Google Gemini mixed up some apples, with other varieties, except for Granny Smith, which we think because it has a distinct colour and shape. It also mixed up a plantain with a banana, since it may look like an unripe banana. Otherwise, it identified everything else, but it still said that a Navel orange is just an “orange”. Google Gemini got 8/12 correct. Logmeal mixed up the oranges, thinking that a clementine and mandarin orange are tangerines. We think that is because they have that similar small size and flatness. It also said that the plantain was a banana. We believe that it said that because Logmeal would think it looks like an unripe banana. Logmeal got 9/12. We think that with better training, Google Gemini could identify different varieties of apples.

Percentage of correct pictures: Gemini - 66.67% Logmeal - 75%

(% calculation: # of correct pictures / total pictures)

 

Variable 3- Clear bag vs Translucent bag vs No bag:

Google Gemini seems to not be affected by the bags, because its answers are the same vs different bags and no bags. It identified Ambrosia apples and McIntosh apples wrong, thinking they were Honeycrisp and Braeburn apples. However, we don’t think it was because of the bags, since it said the same thing to the same fruit without a bag. Google Gemini also thought a nectarine was a red apple in a bag, but without a bag, it said it was a nectarine. We think that the bag might have blurred some key details in colours and patterns on the skin of the nectarine. Google Gemini got 19/26 correct. Logmeal seemed to be more affected by the bags and it thought a Granny smith was a pear and Red delicious was a strawberry. We think that the bag might have blurred some details. It also mixed up Tommy Atkins, Nectarines and kiwis, and we think because they were mostly in bags. Logmeal got 14/26 correct. Google Gemini could usually identify the fruits correctly in the clear bag, but not so much in the translucent bags. For Logmeal it seemed like it didn’t really matter what type of bag it was; it mixed almost half of the fruit up.

Percentage of correct pictures: Gemini - 73.08% Logmeal - 53.85%

(% calculation: # of correct pictures / total pictures)

 

Variable 4- Angle and Position:

Google Gemini tried to identify the fruits, but it seems the angle of the fruits affected the results. It just got apples wrong. It was able to identify Granny Smith apples and Red Delicious apples, as well as peach, red mango, and bananas and plantains. We think that because of the colour and shape of the apples- how they are so similar- it made Google Gemini mix them up, but other fruits and Granny Smith and Red Delicious apples have a distinct colour and shape. It didn’t matter what angle the fruit was at, it still mixed up the apples. Google Gemini got 16/25 correct. Logmeal, since it doesn’t identify variety, just said that all angles and types of apples were apples. Since it doesn’t identify variety, we couldn’t tell if the angle affected it. The only ones it got truly wrong were the mangos (thinking it was a mango, then pomegranate), peaches (thinking they were nectarines), and the plantain (thinking it was a banana- though Logmeal has never identified it correctly). It doesn’t seem like the angle affected Logmeal, as it still identified certain fruits. Logmeal got 19/25 correct.

Percentage of correct pictures: Gemini - 64% Logmeal - 76%

(% calculation: # of correct pictures / total pictures)

 

Variable 5- AI software:

Comparing Google Gemini and Logmeal, in short, Google Gemini is much better than Logmeal. Google Gemeni tries to identify the type and variety of fruit, while Logmeal just identifies the type of fruit- though that is what it was made for (it said on the website). Google Gemini generally identifies all the fruits correctly, and usually mixes up certain varieties of apples, such as Ambrosia, Cosmic Crisp, and McIntosh. It sometimes gets oranges wrong, as well as plantains, but can identify all the other things (besides apples). Logmeal can usually identify mangos right, and bananas, and in addition, can also identify that an “apple is an apple”, though it only says “apple”. It also mixes up oranges, peaches, plantains, kiwis. So taking all this into consideration, Google Gemini would probably work best if this was in real life. 

Percentage of correct pictures: Gemini - 75.94% Logmeal - 69.07%

 (% calculation: average # of correct pictures for each variable)

 

 

 

Answering our hypothesis, which was “We think that AI will be able to identify some common and obvious types of fruit, but not different varieties of the same fruit”, our two AI softwares could identify fruit, but some different varieties and/or fruit types messed them up. Our data partially supports our hypothesis. 

Our experiment was to see if AI can identify fruit types and varieties, but with variables that mimic a realistic grocery store. We took pictures of fruits, with these variables, and imputed each picture into each AI software. We chose these certain variables because we wanted this experiment to be as realistic as possible, so in the future, if AI was to be in a store, this information, that we have gathered, could help with future AI development. 

Overall Google Gemini can identify most types and varieties of fruits, though it doesn’t usually identify Cosmic crisp apples, Ambrosia apples, and McIntosh. Sometimes it does not identify plantains, and sometimes it does. 

Logmeal cannot identify varieties of apples, or any fruit- that is not what it was made for. But for other fruits it does a pretty good job of identifying, except for kiwis and peaches. Sometimes it can identify a kiwi, though not usually. Peaches, however, it cannot identify- it just says “nectarine”, maybe because they look similar. 

As for the variables, Google Gemini was most affected by the bags, but the angle did not affect it. Logmeal didn’t seem to really be affected by our variables. Using this information, Google Gemini is much better at identifying fruit types and varieties. With some training and programming it could be better at identifying fruits, and could even eventually be put in use in stores, helping with inventory checks and being more efficient. Overall Google Gemini has a percentage of 75.94% accuracy. Logmeal has a 69.07% accuracy. Though Google Gemini had a higher percentage, Logmeal isn’t far behind, and for some variables had more correct results than Google Gemini. (% calculation: average # of correct pictures for each variable)

We think that Google Gemini could be used in stores for inventory checks, making people's lives easier and more efficient, while also reducing the need for plastic fruit stickers that keep ending up in landfills. Logmeal could too, with more time and work into identifying varieties of fruits and not identifying other objects.

In the end, we think with programming and training, programmers could use this information to update the AI, and with time, there might be an AI programmed just for this purpose: Recognizing fruits at the grocery checkout.

The results that we collect can be useful for identifying fruits with AI, and also Artificial Intelligence image recognition could be used in stores soon, and that would help with inventory check and just make people's lives easier, cheaper, and more efficient. This could also help improve image recognition- they could program it better. Also if this decides to be put in stores, there could be just one AI software that can identify produce, snacks, or anything, with the right programming.

 

1. Due to money, we could not buy more fruits and varieties, which could have provided a larger dataset to perform the experiment and possibly provide more information
2. Unable to access certain AI platforms, or paid ones, which could have been better at identifying types and varieties of fruit
3. It was not possible to setup an environment at home that is exactly the same as a grocery store check-out scanner, and we cannot perform the experiment in an actual store
4. If we used a higher quality and better resolution camera than the one from a smartphone, the images might be better and the AI might have a higher chance of identifying fruits

Boesch, G. (October 10, 2024). Image Recognition: The Basics and Use Cases (2025 Guide). VisoAI

https://viso.ai/computer-vision/image-recognition/

 

Bouffard, K. (June 29, 2012). Lakeland Inventor’s Lasers Could Revolutionize Labeling of Fresh Foods. TheLedger

https://www.theledger.com/story/news/2012/06/30/lakeland-inventors-lasers-could-revolutionize-labeling-of-fresh-foods/26501241007/#:~:text=The%20machine%20uses%20infrared%20laser,Education%20Center%20in%20Lake%20Alfred.

 

Dormer, D. (March 12, 2018). What those little stickers on fruits and vegetables are for. CBC.

https://www.cbc.ca/news/canada/calgary/calgary-plu-fruit-vegetable-sticker-1.4573302#:~:text=Can%20you%20eat%20the%20 stickers,it%20would%20be%20Health%20Canada.

 

Doshi, S. & O’Neal, M. (September 23, 2024). Here’s What to do With Those Annoying Produce Stickers. EcoEnclose.

https://www.ecoenclose.com/blog/heres-what-to-do-with-those-annoying-produce-stickers/#:~:text=For%20now%2C%20PLU%20stickers%20are,the%20shipper%20to%20the%20retailer.

 

Elevate Packaging. (August 23, 2023). Are Produce Stickers Biodegradable? ElevatePackaging.

https://elevatepackaging.com/blog/are-produce-stickers-biodegradable/#:~:text=With%20compostable%20stickers%2C%20you%20can,are%20both%20biodegradable%20and%20compostable.

 

Elevate Packaging. (June 23, 2020). Sustainable Packaging Guide - Eco-friendly Label Adhesive. ElevatePackaging.

https://elevatepackaging.com/blog/ecofriendly-label-adhesive/

 

Elevate Packaging. (n.d.). What Makes Our Bags and Labels Compostable? ElevatePackaging.

https://elevatepackaging.com/composting-standards/

 

Estabrook, R. (November 29, 2011). A Dissolving Fruit Sticker That Claims Soap Superpowers. NPR.

https://www.npr.org/sections/thesalt/2011/11/29/142895565/a-dissolving-fruit-sticker-that-claims-super-soap-powers

 

Gemini (n.d.). Google.

 

IFPS Global. (n.d.) PLU Codes. IFPSGlobal.

https://www.ifpsglobal.com/plu-codes

 

Impress Vinyl. (November 16, 2022). How Long do Vinyl Records Last? How to Extend its Lifespan. VinylPressing.

https://vinylpressing.com.au/blog/vinyl-pressing/how-long-do-vinyl-records-last-how-to-extend-its-lifespan/#:~:text=There%20are%20several%20variables%20that,take%20up%20to%201000%20years.

 

Kachook, O. (May 5, 2021). Produce Stickers: A Small But Mighty Problem. SustainablePackaging.

https://sustainablepackaging.org/2021/05/05/produce-stickers-a-small-but-mighty-problem/

 

Kachook, O. (May 12, 2021). Produce Stickers: Are They The Next Straw? SustainablePackaging.

https://sustainablepackaging.org/2021/05/12/produce-stickers-are-they-the-next-straw/

 

Kachook, O. (May 19, 2021). Produce Stickers: The Benefits Of Going Compostable. SustainablePackaging.

https://sustainablepackaging.org/2021/05/19/produce-stickers-the-benefits-of-going-compostable/

 

Logmeal API. (n.d.)

 

Ohwovoriole, T. (June 27, 2018). What’s The Difference: Biodegradable and Compostable. NaturesPath.

https://naturespath.com/en-ca/blogs/posts/whats-difference-biodegradable-compostable#:~:text=Although%20biodegradable%20materials%20return%20to,but%20with%20an%20added%20benefit.

 

Rayze, G. (n.d.). PLU Finder. PLUFinder.

https://plufinder.com

 

Situ Biosciences. (n.d.). ASTM D6400 – Compostable, Product Test – Composting. SituBiosciences.

https://www.situbiosciences.com/product/astm-d6400-compostable-produc-test-composting/

 

Stock, P. (July 31, 2023). Lasered fruit labels could replace pesky plastic stickers. CosmosMagazine.

https://cosmosmagazine.com/earth/sustainability/lasered-fruit-labels-could-replace-pesky-plastic-stickers/

 

Tucker, B. (November 3, 2015). The truth about fruit stickers. Dirt-mag.

https://www.dirt-mag.com/stories/the-truth-about-fruit-stickers-JWDM20151103151109996

 

Picture Sources:

Nosowitz, D. (March 15, 2018). Those Little Produce Stickers? They’re a Big Waste Problem. ModernFarmer.

https://modernfarmer.com/2018/03/little-produce-stickers-are-big-waste-problem/

 

 

 

Through this project, we had lots of help and received advice from many people. We would like to start by thanking our wonderful parents for helping us in times of need, and giving us advice. We’d also like to thank our teacher, Heather Lai, and Westmount Charter Mid-High, who arranged and planned the school Science Fair. We would like to acknowledge CYSF staff for hosting this event, and all the judges who judged and will judge our project. We would also like to thank all the sources and websites that we used for our research: Viso AI, The Ledger, CBC, EcoEnclose, Elevate Packaging, NPR, Google, IFPS Global, Vinyl Pressing, Sustainable Packaging, Nature's Path, PLU Finder, Situ Biosciences, Cosmos Magazine, Dirt-mag, and Modern Farmer. Thank you so much to everyone for everything that they did to help us through this project!