liveBook · Manning

Bounding box

A bounding box is a coordinate-defined rectangular region in two or more dimensions that encloses or localizes an object, area, or set of points.

How 12 Manning books explain it

Advanced Algorithms and Data Structures

A bounding box is an axis-aligned region that contains a geometric object or curve section.

How it is used
Screens impossible segment and curve intersections and bounds point groups for spatial search.
How deep it goes
Develops containment invariants, insertion choices, intersection screening, and curve subdivision; leaves R-tree split heuristics at a high level.
Examples
Uses overlapping R-tree rectangles, nonintersecting segments with overlapping boxes, and subdivided quadratic Bézier curves.
What you can do after
Apply bounding boxes to reject impossible intersections and understand how spatial trees bound groups of points.

Read how Advanced Algorithms and Data Structures explains bounding box →

Inside Deep Learning: Math, Algorithms, Models

A rectangular region defined by coordinates that encloses an object in an image for localization or annotation.

How it is used
Locates and labels individual objects in images without requiring pixel-by-pixel segmentation labels.
How deep it goes
Defines box extent and implements target tensors, plotting, and PyTorch detector inputs; defers implementing Faster R-CNN from scratch.
Examples
Uses a stop-sign annotation and randomly placed MNIST digits with target and predicted boxes.
What you can do after
Apply it via PyTorch Faster R-CNN by supplying box-and-label targets and inspecting scored box predictions.

Read how Inside Deep Learning: Math, Algorithms, Models explains bounding box →

Fight Fraud with Machine Learning

A bounding box is a rectangular region represented by coordinates that localizes an object or annotated area in an image.

How it is used
Localizes receipt forgery regions for segmentation targets, prediction inspection, and auxiliary text labeling.
How deep it goes
Works through coordinate-to-mask conversion and heatmap-to-box extraction, without detailing the detector architecture.
Examples
Uses forged receipt fields, binary masks, prediction heatmaps, and expert-detected text rectangles.
What you can do after
Apply it via PyTorch and image-drawing utilities to create segmentation labels and inspect receipt-forgery predictions.

Read how Fight Fraud with Machine Learning explains bounding box →

Math and Architectures of Deep Learning

The numeric values that specify the rectangular extent of a detected object within an image.

How it is used
Describes an object’s image location and supplies localization targets for region-based object detectors.
How deep it goes
Defines corner-coordinate representation and connects it to feature-map scaling, regression targets, and detector inference.
Examples
Uses a 256-by-256 RoI mapped to a 16-by-16 feature-map region and examples of multiple localized animals.
What you can do after
Apply via PyTorch-based Fast R-CNN components to represent, regress, and select object-localization coordinates.

Read how Math and Architectures of Deep Learning explains bounding box →

Math for Web Design

A bounding box is a rectangular region described by an element's position and dimensions that encloses the element.

How it is used
Positions one UI element relative to another using their measured rectangles.
How deep it goes
Defines the rectangle, details browser-returned coordinates and dimensions, and works through tooltip positioning; does not explore measurement internals.
Examples
Uses a hover tooltip centered above a button with an 8px gap.
What you can do after
Apply it via getBoundingClientRect() to measure elements and position a tooltip above its button.

Read how Math for Web Design explains bounding box →

Deep Learning with Python, Third Edition

A rectangle representation used to localize an object in an image, typically encoded by coordinates and size.

How it is used
Supports object localization, counting, tracking, cropping, detection training, and prompts for segmentation models.
How deep it goes
Explains corner, pixel, normalized, and grid-relative encodings with preprocessing code, but defers detector architecture details.
Examples
Uses a mango box prompt, COCO cat-bench-bicycle annotations, visualization helpers, and YOLO grid targets.
What you can do after
Build it from scratch by parsing, normalizing, visualizing, and converting box annotations into YOLO training targets.

Read how Deep Learning with Python, Third Edition explains bounding box →

Build AI Drug Discovery Pipelines

A binding box is a three-dimensional spatial region that constrains where a docking algorithm searches for ligand binding poses.

How it is used
Constrains molecular-docking searches to a biologically relevant protein-binding region.
How deep it goes
Explains coordinate-based construction and sizing trade-offs, but stops short of detailed Point and Box implementations.
Examples
Uses a crystallized ligand, coordinate extrema, five-unit padding, and a Python Point–Box implementation.
What you can do after
Operate/configure a docking workflow by deriving a search box from ligand coordinates.

Read how Build AI Drug Discovery Pipelines explains bounding box →

Just Use Postgres!

A bounding rectangle is a rectangular region that encloses the spatial extent of a geometric object or set of objects.

How it is used
Supports efficient geospatial searches by filtering candidate geometries through indexed spatial extents.
How deep it goes
Defines rectangles and explains hierarchical indexing, but stops short of geometric construction or query syntax.
Examples
Uses points, polygons, line segments, hierarchical GiST regions, and SP-GiST quadrant boxes.
What you can do after
Evaluate how spatial indexes narrow searches by traversing bounding boxes instead of scanning every geometry.

Read how Just Use Postgres! explains bounding box →

Human-in-the-Loop Machine Learning: Active learning and annotation for human-centered AI

A rectangular region specified around an object in an image or video frame, commonly used as a visual annotation.

How it is used
Uses bounding boxes to annotate objects and establish ground truth for object-detection tasks.
How deep it goes
Explains expert review, ambiguity margins, and IoU measurement without detailing how boxes are drawn.
Examples
Uses pedestrian detection, expert correction of nonexpert boxes, and boundary cases involving occlusion or objects outside the frame.
What you can do after
Evaluate annotation location accuracy by comparing predicted and ground-truth boxes with intersection over union.

Read how Human-in-the-Loop Machine Learning: Active learning and annotation for human-centered AI explains bounding box →

Designing Deep Learning Systems: A guide for software engineers

A rectangular coordinate representation that localizes an object within an image.

How it is used
Supports object-detection dataset preparation by marking target objects before model training.
How deep it goes
Shows bounding boxes in a labeling workflow but does not specify coordinate formats or geometric implementation.
Examples
Uses a labeling interface where users draw boxes on images before automated training.
What you can do after
Apply it via a labeling application to mark objects and produce training labels.

Read how Designing Deep Learning Systems: A guide for software engineers explains bounding box →

Hugging Face in Action

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

How it is used
Locates and visualizes objects detected in images or video.
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 you can do after
Apply it via Hugging Face Transformers and PIL to inspect detections and annotate images with labeled rectangles.

Read how Hugging Face in Action explains bounding box →

HTML5 in Action

A rectangular region that encloses a graphic or group of graphics.

How it is used
Uses a graphic group's bounds to support wall checks and firing positions in an SVG game.
How deep it goes
Explains getBBox() with a returned-object example and game uses, while noting browser-specific limitations.
Examples
Demonstrates bounding boxes with a moving, shooting flock of UFOs.
What you can do after
Apply SVG getBBox() to obtain a group's bounds for movement checks and firing positions.

Read how HTML5 in Action explains bounding box →

Also covered in

What to know first

Related concepts

Questions about bounding box

What is bounding box?

A bounding box is a coordinate-defined rectangular region in two or more dimensions that encloses or localizes an object, area, or set of points.

Which Manning books explain bounding box?

Advanced Algorithms and Data Structures, Inside Deep Learning: Math, Algorithms, Models, Fight Fraud with Machine Learning, Math and Architectures of Deep Learning, Math for Web Design, Deep Learning with Python, Third Edition, Build AI Drug Discovery Pipelines, Just Use Postgres! and 4 more.

What should I know before learning bounding box?

Manning books that teach bounding box usually introduce coordinate, language model, image segmentation and pixel first.

What kinds of bounding box are there?

Kinds of bounding box include geo-bounding box.

These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.