Image Classification

Image · HuggingFace · Visual

The Image Classification detector runs any HuggingFace image-classification pipeline on image assets. It returns labels and confidence scores, optionally maps specific labels to severity levels, and stores findings alongside the source image reference.

When to use

NSFW detection — classify uploaded or scanned images for adult or explicit content before they reach storage or downstream systems.
Harmful content filtering — flag images containing violence, self-harm indicators, or other policy-violating categories.
Custom image category labelling — deploy a model fine-tuned on your own image dataset to tag product photos, document scans, or media assets.
Automated content moderation — route high-severity predictions to a human review queue using severity_map.
Compliance and data governance — detect sensitive image types (e.g. ID documents, medical imagery) in data lake scans.

How it works

Classifyre loads the HuggingFace vision model and runs the image-classification pipeline on each image asset. The top-k predictions above confidence_threshold are stored as findings. Labels in the severity_map override the default severity level.

The default model is google/vit-base-patch16-224 when no model is specified — a general-purpose ViT that serves as a useful baseline before you deploy a task-specific checkpoint.

Configuration

Parameter	Type	Required	Description	Default	Constraints
model	string	No	HuggingFace hub ID or local path. Defaults to 'google/vit-base-patch16-224' when null.	null	—
model_revision	string	No	Git branch, tag, or commit hash when fetching from the HuggingFace hub.	null	—
device	string	No	Inference device: 'cpu' (default), 'cuda', 'mps', or a CUDA device string like 'cuda:0'.	cpu	—
top_k	integer	No	Maximum number of top predictions to return per image.	null	—
function_to_apply	string	No	Score normalization applied after the model forward pass.	null	—
confidence_threshold	number	No	Minimum prediction confidence to report a label as a finding (0-1). Defaults to 0 so all top_k predictions are reported.	0	min 0, max 1
severity_map	array	No	Ordered rules mapping predicted labels to severity levels. Labels with no matching rule receive 'info' severity.	null	—

Examples

NSFW image detection (default model)

{
  "type": "IMAGE_CLASSIFICATION",
  "confidence_threshold": 0.75,
  "severity_map": [
    { "label": "nsfw", "severity": "high" },
    { "label": "suggestive", "severity": "medium" }
  ]
}

Custom document-type classifier

{
  "type": "IMAGE_CLASSIFICATION",
  "model": "my-org/document-type-classifier",
  "model_revision": "main",
  "device": "cpu",
  "top_k": 3,
  "confidence_threshold": 0.6,
  "severity_map": [
    { "label": "passport", "severity": "high" },
    { "label": "id_card", "severity": "high" },
    { "label": "medical_record", "severity": "high" }
  ],
  "severity": "info"
}