👁️

Object Detection - DETR

🛠

Find Everything in Your Image

Detects and localizes multiple objects in photos, returning bounding boxes and confidence scores. Uses transformer architecture (ResNet-50 backbone) to identify what's in your image and where.

What it does

DETR (DEtection TRansformer) finds and identifies objects in images. Point it at any photo, and it returns a list of what it found with precise locations (bounding boxes) and confidence scores. Instead of scanning the image piece-by-piece like older methods, DETR uses transformers to understand the entire image at once and directly predict all objects. Can detect 80+ object types: people, cars, dogs, chairs, cups, backpacks, and much more.

Problem it solves

  • Detect multiple objects in a single image (people, animals, vehicles, furniture, etc.)
  • Localize objects with precise bounding boxes showing where each object is
  • Real-time object detection for video surveillance, monitoring, autonomous systems
  • Visual understanding for image search, content moderation, accessibility
  • End-to-end pipeline fully trainable (no hand-crafted region proposals)
  • General-purpose detection across 80+ object categories
  • Batch processing for efficient large-scale image analysis

Input/Output

  • Input: RGB image
  • image
  • Output: List of detected objects
    • Class label: What the object is (e.g., "person", "dog", "car", "chair")
    • Bounding box: Location in image [x_min, y_min, x_max, y_max]
    • Confidence score: How sure (0.0–1.0, where 1.0 = 100% sure)
    • image

Accuracy

  • Quality by Object Type:
    • Large, distinct objects: Excellent detection (90%+ confidence)
    • Medium objects: Very good detection (70-85% confidence)
    • Small objects: Reasonable detection (50-70% confidence)
    • Occluded objects: May miss or detect as separate objects
    • Crowded scenes: Good, handles multiple overlapping objects

Technical Details

Architecture
Encoder-Decoder Transformer (DETR)
Core Innovation
Treats object detection as direct set prediction instead of region-based detection. Uses 100 learnable "object queries" that simultaneously search for objects and predict their classes and locations.
Components
• CNN Backbone (ResNet-50): Extracts multi-scale image features • Positional Encoding: Encodes spatial coordinates for the transformer • Transformer Encoder: Processes global image context via multi-head attention • Transformer Decoder: 100 object queries with cross-attention to encoder features • Prediction Heads: Classification (80 COCO classes) and bounding box regression (x, y, w, h)
Training
• Dataset: COCO 2017 (80 object categories) • Loss: Bipartite matching (Hungarian algorithm) + cross-entropy classification + L1 + Generalized IoU localization • Optimizer: AdamW over 150 epochs with learning rate decay
Requirements
• PyTorch (v1.7+) • GPU recommended for inference

Compliance & Provenance

Provider
Open-source
Provider type
Specialized
License
Apache 2.0
EU AI Act risk class
Minimal Risk
Art. 50 transparency
Not applicable
Region availability
Available globally
Training data summary
Pending — provider has not yet published per Art. 53(d)

For more on how we classify models and mark outputs, see our AI Policy.

Model Source