Automated imaging tagging for IOT startups
Use image classification to easily tag photos your devices take, using whatever rules you want.
"With Nyckel, building ML models takes just minutes, which lets me quickly iterate and build exactly what I want."
Get real-time insights about the photos your smart devices take
Set up rules for how the images should be tagged, and then classify them with our API.
See one in action for a smart doorbell
Is there a package at the door?
Upload a picture yourself, or drag one of the ones below.
Drag & drop an image to test
Create custom classifiers or use a pretrained one
Whether it's out-of-the-box or specific to your business, building a classification model takes just minutes.
Say goodbye to idle GPU time. Pay for requests instead.
Since your device could be dormant for hours at at time, request pricing makes more sense than per-hour "always-on" pricing.
Automatic retraining means no manual deployment.
The model auto-retrains after every change, so you never have to manually deploy your models.
Monitor what people are uploading in real-time.
Get a pulse on what images are being uploaded, making it easy to see how your device is being used.
Use the API to create models, add samples, and make requests
Ping us, and we'll respond with the predicted label and confidence score.
How it works
Nyckel abstracts the backend work so you can focus on your core product.
Connects with 100+
Deep Neural Networks
Ensures these tagging models have high accuracy.
Streamlines data
processing for you
Train models with as few as 2 samples per class.
Scales from day 1
thanks to elastic requests
Pay only for what you use - not for idle GPU time.
Classification problems deserve a classification solution.
| Other classification tools |  |
|
|---|---|---|
| Pricing model | Confusing pay-per-hour GPU usage | Per request pricing. Models are free to set live |
| Training | Takes hours to days | Takes 30-60s, providing instant accuracy feedback |
| Accuracy | Adequate | Class-leading thanks to 100s of latest models and fine-tuning techniques |
| Samples needed | 100s-1000s | Start with 2 per class |
| Deployment | Manual, using hardware you hope is optimal, with crude scaling control | Automatic, straight to optimized hardware with per-request auto scaling |
| Model improvement | Manually retrain with new data | Automatically retrains with new data and models |
Build custom classifiers for your IOT device today
Build, train, and deploy classifiers through our IOT classification API. And never pay for idle GPU time - only for actual request usage.