Bloom logoBloom
Agents/image-classifier

image-classifier

v0.3.1

Returns labels, confidence scores, and detected objects for any image URL or base64 input. Supports custom model fine-tuning.

Install

bloom install image-classifier@0.3.1
Stars2
Downloads4537
Version0.3.1
PublishedMar 4, 2026

README

Overview

image-classifier runs multi-label image classification, returning confidence scores for detected labels and bounding boxes for object detection.

Usage

{
  "url": "https://example.com/photo.jpg",
  "topK": 5,
  "model": "general-v2",
  "detect": ["objects", "faces", "text"]
}

Output

{
  "labels": [
    { "name": "outdoor", "confidence": 0.97 },
    { "name": "mountain", "confidence": 0.94 },
    { "name": "snow", "confidence": 0.88 }
  ],
  "objects": [
    { "name": "person", "confidence": 0.91, "box": [120, 45, 210, 380] }
  ],
  "text": []
}

Models

  • general-v2 — 10,000+ label taxonomy
  • nsfw — Content moderation
  • faces — Face detection and attributes
  • Custom fine-tuned models via model ID

License

MIT

Agent Schema

{
  "name": "image-classifier",
  "inputs": {
    "type": "object",
    "properties": {
      "url": {
        "type": "string",
        "format": "uri"
      },
      "topK": {
        "type": "number",
        "default": 5
      },
      "model": {
        "type": "string",
        "default": "general-v2"
      },
      "base64": {
        "type": "string"
      },
      "detect": {
        "type": "array",
        "items": {
          "enum": [
            "objects",
            "faces",
            "text",
            "labels"
          ],
          "type": "string"
        }
      }
    }
  },
  "outputs": {
    "type": "object",
    "properties": {
      "text": {
        "type": "array"
      },
      "faces": {
        "type": "array"
      },
      "labels": {
        "type": "array"
      },
      "objects": {
        "type": "array"
      }
    }
  },
  "runtime": "python",
  "version": "0.3.1",
  "description": "Returns labels and confidence scores for an image",
  "capabilities": [
    "image-classification",
    "object-detection",
    "vision"
  ]
}