Fridgr.

Inspiration

Consumerism and Food Wastage are some of the main drivers of major global crises such as Climate Change, Pollution and Global Hunger. Such issues are mainly fueled by wasteful food habits, such as impulse buying and a lack of clarity on expiration dates. Uncertainty around expiration dates leads to about 20% of food waste in households (U.S. Food & Drug Administration, 2019), whereas excessive buying, which includes unplanned purchases, moreso impulse purchases, is also one of the main factors associated with higher consumer food waste (Lee, 2018, Setti et al., 2018, Stancu et al., 2016) cultivating million tonnes of food wasted each year. Hence, the ultimate goal of this project is to bring a minimalistic yet modern solution to such persistent concerns, to design and build a smart fridge system, where the bulk of our food items are stored, that eventually promotes greater convenience and efficiency in the chore of grocery shopping — Fridgr.

What it does

The main features of Fridgr. includes :

A tracker which tracks the category, barcode, and expiry date of the food placed in the fridge and stores it in a database.
A notification system which notifies the users when the food is about to expire

How we built it

Smart Inventory Log

Using a Raspberry Pi Camera V2, a quick scan of the food items, on the side with barcode and expiry date (if applicable) before storage, would be sufficient to input the item information into the household’s personal encrypted database. This is possible with the help of OpenCV’s AI Recognition which allows autonomous identification of the item. The item information is then logged along with its expiry date and serial number. When items are removed from the fridge, a similar action would trigger the erasement of the items’ information from the database. Information from the database is conveniently retrievable by users through a mobile app, allowing informed decisions when grocery shopping and effectively resists impulse buying.

Automated Expiry Data Tracking

Aside from consumerism, the digitalisation of we ease consumers from the burden of consistently checking and memorising expiring dates of items. Our in-built reminder would remind consumers 1 month before, 1 week before and 1 day before the expiry date. As such, unnecessary waste could be cut down by a large margin. In fact, on the date of expiry, consumers would be nudged for disposal of expired items to maintain a hygienic environment within the fridge.

Challenges we ran into

Raspberry Pi 4 cannot display its desktop as we do not have access to monitor. We ended up connecting to it via ssh and vnc.
There are limitations to current computer vision technology. Even with OpenCV, one of the most sophisticated A.I. models, we could only accurately recognise item categories and not the actual products themselves. This could render our system ineffective in sorting. However, the A.I. model is able to identify a few possible identity of an item through image search. Hence, a potential workaround would be to prompt the user to select the correct item name after every input is registered, which could compromise user experience.
For the current model, we are hosting the server on the raspberry pi, which only works if the user is connected to the local network as the raspberry pi. If we need to access it outside the local network, we’ll need to host the server elsewhere or port forward the raspberry pi.
Preparation and confirmation for supplies provided in a hackathon would allow much smoother execution of our ideas. Going to a software-centric hackathon to create a hardware hack was already hard enough, but not having the right components and having to compromise in various aspects definitely hurt. A much better outcome could be achieved in future projects with ample preparation and communication.

Accomplishments that we're proud of

We had many other ideas in mind before coming to this competition. However, when faced with very little hardware to work with (1 Raspberry Pi 4 and 1 Raspberry Computer V2 only), we immediately came up an original and novel idea on the spot and was able to execute and complete it in terms of hardware and software.
We incorporated A.I. in many aspects of our project, including execution of our project and within our product itself. We generated basecodes and debugged using A.I. which saved us a lot of time coding and were able to better focus on technical details and creative works.
We realised the concept Internet of Things (IoT) by connecting our fridge camera to a Raspberry Pi 4 which communicates with our mobile app to form a dynamic smart fridge system.

What's next for Fridgr.

Fridgr would require lower costs and more efficient memory management in order to scale-up. We would reconstruct the physical design such that it would cost less to build and operate. A new memory system would also be configured to allow more efficient input and retrieval of data.

Stay tuned for Fridgr 2.0!