IRISAI Smart Mirror
A mirror that knows what suits each shopper, and keeps what it sees to itself.
IRIS is an on-device computer-vision mirror built on the NVIDIA Jetson Nano Developer Kit. It runs face and attribute detection and virtual try-on at the edge, then recommends products from your catalogue. Every shopper gets a personal experience, and no video ever leaves the device.
- Runs on
- NVIDIA Jetson Nano inside the mirror
- Input
- Camera facing the shopper
- Output
- Virtual try-on and recommendations
- Privacy
- No video leaves the device
The mirror is where buying decisions happen.
Shoppers stand in front of a mirror at the exact moment they decide whether something suits them. In most stores that mirror is just glass. There's no help choosing, no idea of what else would work, and no way to see another option without another trip to the rack.
IRIS turns that moment into a personal styling session. A camera and an NVIDIA Jetson Nano inside the mirror detect the shopper's face and style-relevant attributes, overlay products on the live reflection and suggest what to try next. All processing happens inside the mirror, so there is no footage to store, secure or send.
What changes with IRIS
- Plain glassLive try-on on the shopper's own reflection
- GuessworkRecommendations matched to the shopper's attributes
- Missed add-onsCross-sell suggestions while the shopper is still deciding
- Privacy worriesNo camera footage ever leaves the mirror
How IRIS works
Everything below runs on the Jetson Nano in the mirror, frame by frame, without a round trip to the cloud.
Detect
A camera in the mirror frame notices a shopper standing in front of it and locates their face.
Understand
On-device models find facial landmarks and estimate attributes such as face shape and age band, in real time on the Jetson GPU.
Try on
Chosen products are overlaid on the live reflection and follow the shopper's movement, so they can compare options without changing.
Recommend
A recommendation engine on the device matches attributes and the current try-on against your catalogue and shows the best options.
Features
Personalisation that normally needs the cloud, delivered entirely on the device.
Face and landmark detection
Finds the face and key landmarks in each frame, so overlays sit in the right place and follow movement.
Attribute estimation
Estimates style-relevant attributes, such as face shape and age band, to personalise every suggestion.
Virtual try-on
Overlays products such as eyewear and accessories on the live reflection, so shoppers compare options in seconds.
Personal recommendations
Suggests products from your catalogue that suit the shopper and pair with what they are trying.
Cross-sell at the mirror
Shows complementary items while the shopper is still deciding, when they are most open to adding on.
Privacy-safe by design
Frames are processed in memory on the device and discarded. No video or images are uploaded or stored.
Works offline
Detection, try-on and recommendations don't depend on the internet. A connection is only needed to sync the catalogue.
Catalogue sync
Product images and recommendation rules update from your catalogue, so the mirror reflects your current range.
Engagement insights
Optional, anonymous counts of sessions and try-ons, so you know which products get tried most.
Technical specifications
IRIS is a self-contained unit: camera, compute and display in one mirror. The heart of it is NVIDIA's Jetson Nano.
NVIDIA Jetson Nano Developer Kit
- GPU
- 128-core NVIDIA Maxwell
- CPU
- Quad-core ARM Cortex-A57, 1.43 GHz
- Memory
- 4 GB 64-bit LPDDR4
- AI compute
- 472 GFLOPS (FP16)
- Power
- 5 W and 10 W modes
- Compute
- NVIDIA Jetson Nano Developer Kit, inside the mirror
- Inference
- Fully on-device, with models optimised for the Jetson GPU using NVIDIA TensorRT
- Vision models
- Face detection, facial landmarks and attribute estimation
- Try-on
- Real-time overlay of catalogue items on the live camera view
- Recommendations
- On-device ranking against your catalogue, using attributes and the current try-on
- Camera
- Front-facing camera integrated in the mirror frame
- Display
- Mirror finish over a display panel, sized to your store fixture
- Connectivity
- Wi-Fi or Ethernet for catalogue sync; detection and try-on work offline
- Data handling
- Frames processed in memory and discarded; no video or images leave the device
- Analytics
- Optional anonymous engagement counts, such as sessions and try-ons
Better for shoppers, stores and the brand
IRIS makes the mirror useful to everyone who stands in front of it, and to the team that runs the store.
Shoppers
- Try more options in less time
- Get suggestions that actually suit them
- No app, no sign-up and no photos taken away
Store teams
- Fewer trips to the rack and the trial room
- Natural add-on suggestions at the mirror
- A reason for passers-by to step in
Brand and leadership
- Personalisation without collecting personal data
- Learn which products get tried most (optional, anonymous)
- An in-store experience competitors don't have
Use cases
Wherever a shopper looks in a mirror before deciding to buy.
- Eyewear storesTry frames on the reflection and see shapes matched to the face.
- Fashion and apparelStyling suggestions and add-ons at the fitting mirror.
- Jewellery and accessoriesPreview pieces on the reflection with less handling of stock.
- Beauty and cosmeticsProduct suggestions at the counter, matched to the shopper.
- Department storesOne mirror that cross-sells across categories.
- Launches and pop-upsAn interactive mirror that draws a crowd and measures it anonymously.
Questions about IRIS
Can't find your answer here? Ask us directly.
Does IRIS record or upload video?
No. Frames are processed in memory on the Jetson Nano inside the mirror and discarded. No video or images are stored or sent anywhere.
Does it need the internet?
Detection, try-on and recommendations run on the device. A connection is used only to sync your catalogue and, if you enable it, anonymous engagement counts.
Which products can shoppers try on?
Try-on works best for products worn on or near the face and upper body, such as eyewear and accessories. We confirm the categories for your catalogue during scoping.
Can it use our own catalogue?
Yes. IRIS recommends from your product range. We set up the catalogue sync and recommendation rules with your team.
Can the mirror match our store design?
Yes. The mirror's size and finish are chosen to fit your fixture, and the on-screen look follows your brand.
Put IRIS in your store.
Tell us about your category and store format, and we'll suggest a mirror setup.