Play showreel

People take drugs. That won't change.  But can technology help educate them about the risks they are taking?

001 →

Drugs are found throughout all areas of society regardless of race, age or background.
However, one thing is for certain. No one ever really knows what they’re taking.

This is a big problem but one that we were confident technology could help solve.

We took three days away from client work to explore how are where technology could be used for good. To automatically identify a drug and inform someone about what they may be about to take, wherever they are.

We have a simple mission. Can our work save one life?

002 →


Pill-ID is a simple and anonymous digital experience to deliver drug information anywhere in the world.

It's aimed at people who are about to take an MDMA based pill and would like information on what it may contain.

Ease of use is key. You simply open your smartphone camera and take a photo of the pill, or upload an existing image. The photo is analysed by a machine learning algorithm, an online database is referenced, and the results are then shown.

Information includes the likely drug, any public warnings, potency, and expected side effects.

003 →


People won't wait to download an app so we need convenience to be effective.

A QR code is used to trigger the initial experience which can then be bookmarked for future use.

The database 1/3
There are several publicly funded organisations and charities that provide updated information and warnings on drugs via publicly accessible websites and databases.

Pill-ID uses a French database called Nuit-Blanche which provides data on different narcotics including dosage, colour, shape etc.

Whenever the Nuit-Blanche database is updated, Pill-ID can then use this information.
Python scripts use existing Nuit-Blanche datasets to compile available information and images for each pill.

This information is then surfaced within Pill-ID - a Progressive Web App (PWA) built using Next JS.

We have selected MDMA based pills for Pill-ID V1 as they typically have a distinct shape and colour that is used as the identifier.

Machine Learning3/3
Pill-ID uses an image input (either a photograph or saved image) and machine learning to predict which drug type in the dataset is a "high confidence" match.

We’ve currently scanned 600 images and there are 15 pills currently in the dataset we have built.

A machine learning model has been built and trained using Tensorflow.js. The model is used at run time by TensorflowJS in the browser (desktop, mobile) and relevant data is surfaced based on the matched image.

With larger datasets the accuracy of the prediction model increases, and a wider selection of pills could be detected.

What's next?

Pill-ID was created by Rehab and Herezie in just three days. It's a proof of concept and MVP (minimum viable product) that successfully shows how technology can help inform and educate audiences about drug risk levels in a convenient way using their phone.

This is just the start and there are a number of potential use cases for this type technology within several different sectors.
N → 001
Similar technology can be used to help medical businesses to keep their patients safe.

Could an Alzheimer's patient use similar technology to help remember the medicine they have taken that day, or what they still need to take.

Or, could this experience be used by pharmaceutical businesses to help inform patients of dosage levels within OTC, non-perscription medicine at the point of purchase or consumption?
N → 002
Health services
There is a global opportunity for this experience and there are numerous international databases that could be integrated with and referenced to deliver information to users.

Additionally, could Pill-ID be linked into the emergency health services to provide help or medical advice?

Or could the anonymous data be used to spot trends and proactively flag health issues to health services? 
N → 003
Could we integrate with multiple databases in different countries to identify global trends?

Could anonymous data be used by health services  to create more targeted, effective education campaigns relating to drug prevention?

To find out more about the prototype or to receive future updates please enter your details and we’ll get in touch.

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.

Rehabstudio Limited and Herezie Group do not, in any way, condone recreational drug use. The purpose of this tool is explicitly for informational purposes and harm reduction. All data shared is anonymous and untraceable.

Completely accurate identification of drugs cannot be guaranteed and you should not take any drugs based on identification by this tool. The database used is and you can read their full legal notice here. The possession, consumption and trade of most of the substances on this database are illegal (and, as such, you should not take them), and all of the substances should be assumed to carry significant health risks. We are not qualified to express an opinion that you are fit to safely take any drugs. All information and images shared are anonymous and untraceable. In the absence of any negligence or other breach of duty by Rehabstudio Limited or Herezie Group, use of this tool is entirely at your risk.