Datasetpublic

BowerApp/bower-waste-annotations

Dataset Card for waste annotations made by the recycling solution Bower The data offered by Bower (Sugi Group AB) in collaboration with Google.org Dataset Summary The bower-waste-annotations dataset consists of 1440 images of waste and various consumer items taken by consumer phone cameras. The images are annotated with Material type and Object type classes, listed below. The images and annotations has been manually reviewed to ensure correctness. It is… See the full description on the dataset page: https://huggingface.co/datasets/BowerApp/bower-waste-annotations.

sourceHugging Facemitupdated 2y agoView on Hugging Face

4likes260downloads

Dataset Card

Dataset Card for waste annotations made by the recycling solution Bower

<p align="center"> <b>The data offered by Bower (Sugi Group AB) in collaboration with Google.org </b> </p>

Dataset Summary

The bower-waste-annotations dataset consists of 1440 images of waste and various consumer items taken by consumer phone cameras. The images are annotated with Material type and Object type classes, listed below. The images and annotations has been manually reviewed to ensure correctness. It is assumed to have high quality of both bounding box accuracy and material + object combination. The purpose of the dataset has been to validate models - therefor all images in this set is categorized as validation data. As quality of annotations is high, one can use this as ground truth for validating models. This data set was created in a collaboration with Google.org and Google employees as part of the Google Impact Challenge: Tech for social good support that Bower got H1 2024. Which had as goal to create impact in the area of sustainability and waste. Any use of this data that contributes to this purpose is highly appreciated! Let us know how you contribute.

Dataset Description

—Homepage: https://www.getbower.com
—Repository: https://github.com/PantaPasen/seegull
—Point of Contact: lucas@getbower.com

Number of Images

json

{'valid': 1440}

Taxonomy and description

Object type

Object Type	Description
`Aerosol`	A gas-tight, pressure-resistant container with a valve and propellant. When the valve is opened, the propellant forces the product from the container in a fine or coarse spray pattern or stream. Example: spray cans for paint, furniture polish, etc., under pressure. Excludes atomizers, as they do not rely on a pressurized container to propel product.
`Battery`
`Blister pack`	A type of packaging in which the item is secured between a preformed (usually transparent plastic) dome or "bubble" and a paperboard surface or "carrier." Attachment may be by stapling, heat-sealing, gluing, or other means. Often thermoformed from polyvinyl chloride, although other thermoplastics can be used.
`Book`
`Bottle`	A container with a round neck of relatively smaller diameter than the body, capable of holding a closure for retention of contents. It may be round, oval, square, oblong, or a combination, and is generally made of glass or plastic. Bottles can be disposable, recyclable, returnable, or reusable.
`Box/Card/Tray/Crate`	A non-specific term for a rigid, three-dimensional container with closed or semi-closed faces. Boxes and crates are made from various materials and may be used for storage or display. Trays are shallow containers used for display. Includes open-ended non-rigid containers like sleeves.
`Gable Top`	Package for liquids shaped like a gable top (triangle top), usually made of cardboard.
`Aseptic Flat Top Carton`	A rectangular-shaped, stackable package designed for liquids like juice or milk, made of multilayer laminated materials (e.g., Tetra Pak).
`Tube`	A cylindrical container sealed on one end, closed with a cap or dispenser on the other (e.g., toothpaste tube).
`E-cigarette`
`Bulb`
`Can/Tin/Cylinder`	Metallic, generally cylindrical containers of unspecified size, used for various items. Cylinders have rigid, straight sides with circular ends of equal size.
`Cosmetics containers`	Containers for makeup products like eyelashes, eyeliner, lipsticks, etc.
`Cup/Tub/Bowl/Bucket`	Flat-bottomed containers with a base of any shape, usually used for foods like ice cream, margarine, yogurt, and sour cream. They may or may not have lids and are typically made of paper, plastic, or other materials.
`Jar`	Rigid container made of glass, stone, plastic, or other material, with a large opening for storing products such as jams and honey.
`Electronic device`
`Wrapper/Foil/Net/Bag/Flexibles`	Flexible containers and materials used to enclose all or part of an item. Includes wrappers, gas-tight packs, nets for holding items, and bags or other flexible containers that may be sealed.
`Jug`	Cylindrical container with a handle, lid, or spout, used for holding and pouring liquids.
`Lid/Cap`
`Other medical containers/Pharmaceuticals`	Example: Nasal spray containers
`Paper product`	Includes white paper, newspaper, magazines, and other paper-based items.
`Other`

Material type

Material Type	Description
`Hard plastic`	Rigid plastics that maintain their shape, such as bottles, containers, toys, and household items made from materials like PET or HDPE.
`Soft plastic`	Flexible plastics that can be easily bent or scrunched, including plastic bags, wrappers, and films made from materials like LDPE.
`Other plastic`	Plastics that don't fit into hard or soft categories, possibly including bioplastics, mixed polymers, or specialized plastic products.
`Fiber - Paper`	Thin sheets made from pressed fibers, used for writing, printing, and packaging, such as office paper and newspapers.
`Fiber - Cardboard / Paper Packaging`	Thicker paper materials used for packaging, like corrugated cardboard boxes, paperboard cartons, and other sturdy paper packaging.
`Fiber - Other`	Other fiber-based materials not classified as standard paper or cardboard, such as insulation or specialized industrial fibers.
`Metal`	Materials made from metals like aluminum, steel, copper, and other metal alloys, used in cans, appliances, and construction materials.
`Glass - Coloured`	Glass materials that are tinted or colored, such as green or brown bottles and decorative stained glass items.
`Glass - Non-coloured`	Clear or transparent glass materials like window panes, clear bottles, and jars without any coloration.
`Composite`	Materials made from two or more different substances that remain distinct, such as Tetra Pak cartons combining paperboard, plastic, and aluminum.
`Hazardous waste`	Waste materials that are dangerous or potentially harmful to human health or the environment, like chemicals, batteries, and pesticides.
`Electronic waste`	Discarded electronic devices and components, including computers, smartphones, televisions, and electronic accessories.
`Organic waste`	Biodegradable waste derived from living organisms, such as food scraps, yard waste, and agricultural waste.
`Textiles`	Fabrics and cloth materials made from natural or synthetic fibers, used in clothing, upholstery, and linens.
`Wood`	Material derived from trees, including lumber, plywood, and wooden products used in construction and furniture.
`Rubber & Leather`	Materials made from natural or synthetic rubber (like tires) and animal hides (leather goods such as shoes and belts).
`Other`	Materials that do not fit into any of the specified categories above, including unique or miscellaneous items.

Languages

English

Data Instances

A sample from the validation set is provided below:

{
  ...
}

Data Splits

Validation dataset only

Dataset Creation

Images collected from consumer phone cameras and annotated manually

Annotations

Annotation process

1.Pre-labelling using Grounding-DINO model (bounding boxes and labels)
2.Manual annotation using LabelBox
3.Review by Taxonomy expert for ensured quality

Considerations for Using the Data

Appreciate creditaion of Bower (Sugi Group AB) from usage of the data

Social Impact of Dataset

This dataset was created in a collaboration with Google.org and Google employees as part of the Google Impact Challenge: Tech for social good support that Bower got H1 2024. Goal with the project was to increase recycling rates throughout the world - this by incorporating Computer vision solutions in to the Bower-app to enable users to scan any type of trash and get correct sorting guidance on these items. Which had as goal to create impact in the area of sustainability and waste. Any use of this data that contributes to this purpose is highly appreciated! Let us know how you contribute.

Discussion of Biases

Other Known Limitations

Additional Information

Contributors

Lucas Nilsson, Linda Attby, Louise Rönne, Henrik Erskérs, Jeremy, Suhani

Licensing Information

This work is licensed under a MIT license

Contributions

Thanks to Google.org for making this possible, as well as the Google Fellows joining Bower for 6 months to enable this (Jeremy, Vicent, Suhani, Vivek, Ole and Rita).