-
-
v2. cardboard, two IR-sensors, Arduino Nicla Vision. abandoned since we had the ambition to build smth. for your pocket.
-
v3. struggled a lot with making the classification inside work. very limited foldability.
-
v4. fully foldable. the little knob is our sensor: Nicla Vision handles classification & detection. more software, less parts -> design ↑
-
v4 folded. rip to my wallet.
-
first 3d model. we completely abandoned uv light as a way to distinguish cash later
-
the data made available by v4. live exchange rate calculation re. foreign currency cash, location and location-based counterparty estimation
Inspiration
"cash is free" - no, cash is blind.
I use AI to review all my transactions to give me insights into long term spending patterns, forgotten subscriptions, ... It's saved me a lot of money and is the richest data silo to analyze my personal lifestyle. But I couldn't include any of my cash payments, bcs. I missed a way to keep a record.
What it does
Smart wallet detects every banknote that enters or leaves your wallet. V4, the most current realized version uses a camera. It does automatic bookkeeping. It understands foreign currency, considers exchange rates and the data digital transactions
How we built it
We went through a lot of iterations with the goal to fit the product into a pocket, maximize simplicity, get high accuracy and make it useful.
Challenges we ran into
We had to get training and test data for our classification model. >500 images taken from inside the non-rigid wallet and the accuracy goes down if lighting changes, the wallet folds weirdly, ...
What makes this product great
No one has built this yet. Other "smart" wallets aren't smart they just come with GPS. As AI's ability to use data grows, interesting data becomes more valuable for people. We give this data directly to the consumer. We don't see our data as a standalone "walled garden". People should add it to their digital transaction data. It solves a genuine problem with cash. Like with other standout technology it brings utility in unforeseen ways: For example it could help society solve money laundering by understand where fake notes first show up.
What we learned
Foldable, crammed spaces are really hard envs for microelectronics. Boston is beautiful, which we realized on our 5 walks from and to the hardware store.
What's next for Smart Wallet
We realized early in the build that the hardware available to us wasn't the best design choice. We use a camera that needs light and some distance to record the notes. V5 will be built with a small CIS scanner, it will look more like a ridge wallet. This would also address the problem of users putting in multiple notes at once which v4 couldn't solve.
From my hackharvard application:
"At HackHarvard I'd build smth. that helps [...] people understand their finances better. I'd love to build with protocols Americans rely on for payments, banking and credit." We also wanted to build smth. exciting - involving hardware or training models with new frameworks. We did both.
thx org team for org
videolink: https://youtu.be/NUp7MeLcGUU


Log in or sign up for Devpost to join the conversation.