Computer vision visualizer

Object Detection Visualizer

Detect and label common objects in images or live webcam frames directly in your browser with MediaPipe.

Controls

Input mode
Overlay layers

EfficientDet Lite recognizes common COCO objects and runs locally in the browser.

Preview

Upload a scene, use webcam, or inspect the sample guide.
Object detection input

What this visualizer does

It runs MediaPipe ObjectDetector, draws labeled bounding boxes, and groups repeated detections by class and confidence.

What MediaPipe detects

  • Bounding boxes in source-image coordinates.
  • Class labels and confidence scores.
  • Multiple objects and repeated categories.

Common use cases

  • Scene understanding
  • Inventory demos
  • Safety alerts
  • Camera analytics
  • Accessibility

Code examples

import mediapipe as mp

BaseOptions = mp.tasks.BaseOptions
ObjectDetector = mp.tasks.vision.ObjectDetector
ObjectDetectorOptions = mp.tasks.vision.ObjectDetectorOptions

options = ObjectDetectorOptions(
    base_options=BaseOptions(model_asset_path="efficientdet_lite0.tflite"),
    max_results=10,
    score_threshold=0.5
)

with ObjectDetector.create_from_options(options) as detector:
    image = mp.Image.create_from_file("scene.jpg")
    result = detector.detect(image)
    for detection in result.detections:
        print(detection.categories[0].category_name)

How the model works

  1. The image is resized and passed through EfficientDet Lite.
  2. The model proposes boxes and class probabilities.
  3. Overlapping boxes are suppressed.
  4. Threshold and result limits filter the final output.

Controls explanation

Confidence removes weak detections. Maximum results caps the list, while overlay controls independently show boxes, labels, and centers.

FAQ

This general model recognizes common COCO categories. Specialized objects require a compatible custom-trained detector model.