ResNet — Classification AI Model
ResNet labels an image against the 1000 ImageNet categories — animals, vehicles, food, scenes — and returns the top guesses with confidence. Pick 18 for speed, 50 for balance, 101 for accuracy. CPU-capable, sub-second.
What it does
ResNet labels an image against the 1000 ImageNet categories - animals, vehicles, food, scenes - and returns its best guesses with a confidence score on each.
Three depths are available: 18 for speed, 50 for balance and 101 for accuracy, with 50 the default. Residual connections are what let a network this deep train at all, and it classifies on CPU in under a second.
When to use it
- Labelling what is in a picture, out of a thousand categories
- A confidence score to threshold on, rather than one bare label
- Sorting or routing a large batch of images by their subject
- Trading depth against speed: 18, 50 or 101 layers
- Classifying on CPU, without taking up a GPU to do it
Details
- CategoryClassification
- Year2016
- LicenseApache-2.0
Compliance & Provenance
- ProviderOpen-source · Specialized
- EU AI Act RiskMinimal Risk
- Art. 50 TransparencyNot applicable
Inputs & Outputs
- ImageInputimage
Source image for object classification
- TextOutputstring
Classification label and confidence score
Parameters
- versionstringdefault 50options: 18, 50, 101
ResNet depth (18/50/101)
Which ResNet variant to use. '18' — smallest and fastest. '50' (default) — balanced. '101' — deepest and most accurate, slower. Pick 50 for general use; bump to 101 only when accuracy matters more than latency.
Tags
Alternatives in Image Classification
- Adult Content Detection
Classify image safety (safe / nsfw).
- Gender Recognition
Classify pedestrian gender (male/female).