Bounding box

As explained in Hugging Face in Action

A bounding box is a rectangular region specifying the position of an object within an image or video.

A bounding box is a rectangle that locates a detected object within an image or video. Object detection uses boxes alongside object labels and confidence scores. section 4.2

A detection result commonly provides:

• scores: confidence values for detected objects section 4.2.1
• labels: indices mapped through model.config.id2label section 4.2.1
• boxes: coordinates used to draw rectangles on the original image section 4.2.1

Draw detections with DETR

Process an image with facebook/detr-resnet-50, post-process predictions with the image target size and a 0.9 threshold, then draw each returned box with ImageDraw.rectangle(). Add the mapped label near the rectangle. section 4.2.1 section 6.2.1

Figure 4.9 The bounding box for each detected object is drawn on the original image.

View an annotated image with a bounding box drawn around every detected object. section 4.2.1

How Hugging Face in Action covers it

How deep it goes

Explains detection outputs, threshold filtering, coordinate scaling, and rectangle drawing without covering detector internals.

Examples

Uses office imagery plus dog-and-cat imagery, with labels, confidence values, coordinate arrays, and annotated rectangles.

What it assumes

Assumes an image and a pretrained detection model, while demonstrating preprocessing, inference, post-processing, and visualization.

What you can do after

Apply it via Hugging Face Transformers and PIL to inspect detections and annotate images with labeled rectangles.

How it connects

Connects bounding boxes to object detection, image segmentation, confidence scores, pretrained models, and multimodal models.

liveBookmanning.com

Read Hugging Face in Action on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Advanced Algorithms and Data StructuresA bounding box is an axis-aligned region that contains a geometric object or curve section.
• 

  liveBookmanning.com
  Inside Deep Learning: Math, Algorithms, ModelsA rectangular region defined by coordinates that encloses an object in an image for localization or annotation.
• 

  liveBookmanning.com
  Math and Architectures of Deep LearningThe numeric values that specify the rectangular extent of a detected object within an image.
• 

  liveBookmanning.com
  Math for Web DesignA bounding box is a rectangular region described by an element's position and dimensions that encloses the element.
• 

  liveBookmanning.com
  Deep Learning with Python, Third EditionA rectangle representation used to localize an object in an image, typically encoded by coordinates and size.
• 

  liveBookmanning.com
  Just Use Postgres!A bounding rectangle is a rectangular region that encloses the spatial extent of a geometric object or set of objects.
• 

  liveBookmanning.com
  Designing Deep Learning Systems: A guide for software engineersA rectangular coordinate representation that localizes an object within an image.

Compare how 8 books explain bounding box →

Related concepts

• getBoundingClientRectuses · getBoundingClientRect is a DOM method that returns an element's size and position relative to the viewport.
• Intersection over Unionmeasured by · Intersection over Union is the ratio of the intersection area or volume of two regions to the area or volume of their union.
• geo_shape queryused by · An Elasticsearch geo_shape query matches documents whose indexed geographic shapes have a specified spatial relationship to a query shape.
• spatial indexused by · A spatial index is a data structure that organizes spatial objects or regions to accelerate queries involving locations, extents, distances, or geometric relationships.

This explainer was generated by Manning's knowledge graph from the text of Hugging Face in Action. The section links point to the parts of the book it draws on.