Object detection model

As explained in Math and Architectures of Deep Learning

A model that predicts object classes and their spatial locations in an image.

An object detection model assigns class probabilities and bounding-box coordinates to regions of interest, combining classification with spatial localization. section 11.4.3

In Fast R-CNN, a convolutional backbone computes one feature map for the whole image. RoI pooling maps each proposed region onto that feature map and produces a fixed-size representation through max pooling, avoiding a separate CNN pass for every region. section 11.4.3

Fast R-CNN RoI head

A PyTorch RoI head feeds pooled features through fully connected layers, then emits class scores and four-coordinate regression outputs for each class. Training combines cross-entropy classification loss with Smooth L1 regression loss for positive regions; inference selects the most probable class and its corresponding offsets. section 11.4.3

Figure 11.18 A conv feature map with two regions of interest of different sizes. RoI pooling extracts a fixed-sized output feature map (2 × 2 in this image) from each of the RoIs in a single pass via max pooling. This enables us to extract fixed-sized representative feature vectors for each RoI, which are then fed into further classifier and regressor layers for classification and localization.

Compare differently sized regions as they become fixed-sized 2 × 2 feature maps through RoI pooling. section 11.4.3

How Math and Architectures of Deep Learning covers it

How deep it goes

Explains shared feature extraction, RoI pooling, prediction heads, losses, and inference code, stopping short of full detector training.

Examples

Uses a 256 × 256 region, a 16× subsampled feature map, a 2 × 2 pooled output, and PyTorch head, loss, and inference listings.

What it assumes

Assumes convolutional feature maps and proposed regions, while developing how shared features support fixed-size RoI representations.

What you can do after

Builds a Fast R-CNN RoI head, computes its losses, and performs class-specific localization inference in PyTorch.

How it connects

Connects convolutional backbones, region proposal networks, RoI pooling, image classification, and bounding-box regression.

liveBookmanning.com

Read Math and Architectures of Deep Learning on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Machine Learning Platform EngineeringAn object detector is a machine learning model that identifies object instances and estimates their locations in visual input.

Compare how 2 books explain object detection model →

Related concepts

• YOLOincludes · YOLO is a family of single-stage object-detection models that predict object classes and locations directly from an image in one inference pass.
• RetinaNetincludes · RetinaNet is a one-stage object-detection architecture that uses convolutional feature pyramids with separate classification and bounding-box regression heads.
• object detectionused by · Object detection is a computer-vision task that identifies object instances in visual data, assigns them categories, and estimates their locations.
• computer visionused by · Computer vision is the field of computing concerned with using computational methods to extract and interpret information from images, video, and other visual data.
• single-stage detectorincludes · An object-detection architecture that predicts object locations and class labels directly in one integrated model pass without a separate region-proposal stage.

This explainer was generated by Manning's knowledge graph from the text of Math and Architectures of Deep Learning. The section links point to the parts of the book it draws on.