Early-stage startup, building in public

Point your camera.
Know what it's made of.

Resyn is building the fastest way to identify plastic resin type — PET, HDPE, PVC, LDPE, PP, PS, or Other — straight from a live camera feed, so recyclable material stops getting lost to the wrong bin.

PET · #1
0% confidence
soda bottle detected
resin identified
Our mission

Recycling fails at the point of confusion — we're fixing that point.

Most recyclable material doesn't get lost in the truck or the plant. It gets lost at the bin, in the three seconds someone spends guessing whether something is recyclable. Resyn exists to remove that guess — instant, on-device identification, so the right answer is always one scan away.

WHY NOW

Cameras got good enough

On-device computer vision finally runs fast enough to identify material in real time, on hardware people already carry.

WHY US

Built resin-first

We didn't bolt recycling onto a generic object detector — Resyn is purpose-built around the seven resin codes from day one.

WHERE WE'RE HEADED

From prototype to platform

A working scanner today, a trained model and full app on the way — see exactly where we stand in the roadmap below.

How it works

Three steps, one scan.

No barcode, no manual lookup — just hold the item up and let the camera do the reading.

01

Point

Aim your camera at a bottle, tub, or container. Resyn locates it in the frame and locks on.

02

Scan

Hold still for a 3-second scan. Resyn samples several frames — brightness, color, texture, transparency — and votes on the most consistent read.

03

Sort

Get the resin type, its confidence score, and what to do with it. Below 20% confidence, Resyn tells you it isn't sure rather than guessing.

The resin chart

Seven codes, every plastic you own.

The same numbering system stamped into the bottom of your containers — Resyn is trained to recognize all seven.

1
PET

#1 — Polyethylene Terephthalate

Water bottles, soda bottles, juice bottles
2
HDPE

#2 — High-Density Polyethylene

Milk jugs, detergent and shampoo bottles
3
PVC

#3 — Polyvinyl Chloride

Pipe fittings, blister packaging, tubing
4
LDPE

#4 — Low-Density Polyethylene

Grocery bags, bread bags, squeeze bottles
5
PP

#5 — Polypropylene

Yogurt cups, bottle caps, food containers
6
PS

#6 — Polystyrene

Foam cups, disposable cutlery, foam clamshells
7
OTHER

#7 — Mixed & Unmarked

Multi-layer packaging, unmarked or blended plastics
Roadmap

Where we stand today

We're a small team building in public. Here's the real, current state of the product — not a projection.

Shipped

Camera + heuristic classifier

  • Live feed with YOLO object detection locating bottles and cups
  • 3-second multi-frame scan with brightness, saturation, texture & transparency scoring
  • Confidence-weighted voting across frames, with alternative guesses shown
  • Low-confidence scans flagged instead of forcing a guess
Building next

A real trained model

The current classifier is a hand-tuned heuristic — accurate on broad categories, but not on item-level detail. The next phase replaces it with a custom-trained model:

  • Building a labeled photo dataset across all 7 resin classes
  • Annotating in Roboflow, exporting in YOLO format
  • Training a custom YOLO model on real resin images
  • Swapping it into the existing scan pipeline — same camera, same UI
Python OpenCV Ultralytics YOLO Roboflow React (planned) FastAPI (planned)