MuhammadLab
Computer VisionBrowser-basedNo API requiredTensorFlow.jsMobileNetEducational demo

Object Classification Studio

Upload an image or use your webcam to classify objects directly in the browser using a pre-trained machine learning model.

What is object classification?

Predict the main category.

Object classification is a computer vision task where a machine learning model predicts the category of an image or object. If you upload a dog photo, the model may predict “dog”, “golden retriever”, or “Labrador retriever”.

How does it work?

  1. The image is converted into pixels.
  2. Pixels are resized and normalized.
  3. A neural network extracts edges, textures, shapes, and object parts.
  4. The final layer predicts probabilities for classes.
  5. The highest probability becomes the top prediction.

What model are we using?

MobileNet + TensorFlow.js

MobileNet is a lightweight convolutional neural network designed to run efficiently on limited devices such as phones and browsers. That makes it ideal for real-time teaching demos.

This is an object classification demo. It predicts the most likely class of the whole image. It does not locate objects with boxes. For bounding boxes, use an object detection model such as COCO-SSD.
Image preview appears here.

Results and interpretation

Prediction output

Loading
Model status
Loading MobileNet in the browser...
Input type
No input selected
Top prediction
—
Confidence
—
No predictions yet. Upload an image or classify a webcam frame.

Confidence bars

Confidence bars appear after classification.

Educational interpretation: Run classification to see how the model interprets the image.

Limitations

Use predictions as learning outputs.

  • • The model is trained on a fixed set of image categories.
  • • It may not recognize rare, local, medical, scientific, or custom objects.
  • • Blurry, dark, cropped, or multi-object images may confuse the model.
  • • It classifies the whole image; it does not draw bounding boxes.
  • • Predictions are educational outputs, not expert decisions.

Computer vision comparison

Object Classification vs Other Computer Vision Tasks

TaskMain QuestionOutputExample
Object ClassificationWhat is this image/object?Class label + confidenceImage classified as “cat”
Object DetectionWhere are the objects and what are they?Bounding boxes + labelsBox around a cat and a dog
Object IdentificationWhich specific object/person is this?Specific identityThis is Person A
Object VerificationDoes this match the reference?Yes/No or similarity scoreDoes this face match the ID photo?
Image SegmentationWhich pixels belong to each object?Pixel maskExact cat outline

Try This in Class

Student tasks

  • • Upload a clear image with one object.
  • • Upload an image with multiple objects.
  • • Upload a blurry image.
  • • Upload an unusual object.
  • • Compare how the confidence changes.
  • • Discuss why the model may be wrong.
  • • Explain why classification is different from detection.

Discussion questions

  1. Why does the model sometimes predict a very specific label?
  2. What happens when the image contains more than one object?
  3. Why is confidence not the same as truth?
  4. Why is this not object detection?
  5. What dataset would classify custom university lab equipment?
Technical Notes
Library
TensorFlow.js
Model
MobileNet
Execution
Browser/client-side
Input
image element, video frame, or canvas
Output
predicted class names and probabilities
Privacy
image stays in the browser