bounding box
In plain wordsThe rectangle a vision system draws around an object it has found, stored as a handful of numbers giving the rectangle's position and size.
The smallest axis-aligned rectangle enclosing an object in an image: the standard output of object detectionobject detection, n.: Finding every object of interest in a picture or video frame, saying what each one is and drawing a box around it, for example all the cars, cyclists and pedestrians in a street scene. and a common, cheap form of annotationannotation, n.: Adding information to raw data by hand or with software — marking objects in photos, tagging names in sentences, transcribing speech or ranking chatbot answers — so machines can learn from it.. It is stored as four numbers in one of several conventions: corner coordinates (x_min, y_min, x_max, y_max); top-left corner plus width and height, as in COCOCOCO, prop. n.: A widely used collection of everyday photos in which people outlined and named the objects, described each picture in sentences and marked body joints; it is used to train and compare vision systems.; or center plus width and height normalized by image size, as in YOLOYOLO, prop. n.: A family of fast object-detection models that look at an image once and immediately output boxes and labels for everything they recognize, quick enough for live video. label files. Detectors learn box regression, predicting offsets from an anchoranchor box, n.: A ready-made box shape, placed at many spots across an image, that an object detector stretches and shifts to fit real objects instead of drawing boxes from scratch. or reference point, and are scored by the intersection over unionintersection over union, n.: A score from 0 to 1 for how well two shapes overlap, such as a predicted box and the true box around an object: the shared area divided by the combined area. between predicted and ground-truth boxes. Variants include oriented (rotated) boxes for aerial and document images and 3D boxes (cuboids with position, size and heading) for lidarlidar, n.: A sensor that measures distances by firing laser pulses and timing their reflections, building a 3D picture of its surroundings out of millions of points; widely used on robots and self-driving cars. and driving data. Boxes are coarse: thin, diagonal or partly hidden objects fill them poorly, one motivation for pixel masks in instance segmentationinstance segmentation, n.: Finding each separate object in an image and tracing its exact outline, so that two people standing together get two distinct shapes rather than one blob..
Convert the bounding boxes from COCO's corner-plus-size format to normalized center coordinates before training.
Origin
Bounding boxes and bounding volumes were standard in computer graphics and computational geometry by the 1970s–80s, where they speed up clipping, ray tracing and collision tests. In vision they became the default label for detection with benchmarks such as PASCAL VOC (from 2005) and the ImageNet detection track.
Usage note
Often abbreviated 'bbox' in code. Coordinate conventions are a frequent source of silent bugs: check whether a dataset uses pixels or normalized units, and corner or center format.
Related
- object detectionobject detection, n.: Finding every object of interest in a picture or video frame, saying what each one is and drawing a box around it, for example all the cars, cyclists and pedestrians in a street scene.
- intersection over unionintersection over union, n.: A score from 0 to 1 for how well two shapes overlap, such as a predicted box and the true box around an object: the shared area divided by the combined area.
- anchor boxanchor box, n.: A ready-made box shape, placed at many spots across an image, that an object detector stretches and shifts to fit real objects instead of drawing boxes from scratch.
- non-maximum suppressionnon-maximum suppression, n.: A clean-up step in object detection: when the system draws several overlapping boxes around the same object, it keeps the most confident one and throws away the near-duplicates.
- annotationannotation, n.: Adding information to raw data by hand or with software — marking objects in photos, tagging names in sentences, transcribing speech or ranking chatbot answers — so machines can learn from it.
- instance segmentationinstance segmentation, n.: Finding each separate object in an image and tracing its exact outline, so that two people standing together get two distinct shapes rather than one blob.
- instance segmentationinstance segmentation, n.: Finding each separate object in an image and tracing its exact outline, so that two people standing together get two distinct shapes rather than one blob.
Mentioned in
labellabel, n.: The correct answer attached to an example in a dataset — 'cat' on a photo of a cat, 'positive' on an upbeat review — that a learning system is trained to predict., object detectionobject detection, n.: Finding every object of interest in a picture or video frame, saying what each one is and drawing a box around it, for example all the cars, cyclists and pedestrians in a street scene., annotationannotation, n.: Adding information to raw data by hand or with software — marking objects in photos, tagging names in sentences, transcribing speech or ranking chatbot answers — so machines can learn from it., face detectionface detection, n.: Finding where the faces are in a photo or video, without working out whose faces they are: the step that lets a camera focus on people or blur them out., intersection over unionintersection over union, n.: A score from 0 to 1 for how well two shapes overlap, such as a predicted box and the true box around an object: the shared area divided by the combined area., non-maximum suppressionnon-maximum suppression, n.: A clean-up step in object detection: when the system draws several overlapping boxes around the same object, it keeps the most confident one and throws away the near-duplicates., YOLOYOLO, prop. n.: A family of fast object-detection models that look at an image once and immediately output boxes and labels for everything they recognize, quick enough for live video., anchor boxanchor box, n.: A ready-made box shape, placed at many spots across an image, that an object detector stretches and shifts to fit real objects instead of drawing boxes from scratch., visual groundingvisual grounding, n.: Linking words to the exact part of an image they refer to, for example finding 'the man in the red hat on the left' and pointing to him.