16  Deep learning for classification, detection, segmentation and anomaly

Chapter status: Structural placeholder.

This chapter currently contains the approved heading structure but no reviewed technical content.

16.1 Image classification

16.1.1 Whole-image classification pipelines

Content status: Placeholder.

This subsection has not yet been drafted or technically reviewed.

16.1.2 Multi-label classification

Content status: Placeholder.

This subsection has not yet been drafted or technically reviewed.

16.1.3 Fine-grained classification

Content status: Placeholder.

This subsection has not yet been drafted or technically reviewed.

16.1.4 Evaluation with imbalanced industrial datasets

Content status: Placeholder.

This subsection has not yet been drafted or technically reviewed.

16.2 Object detection

16.2.1 Two-stage detectors: R-CNN, Fast R-CNN, Faster R-CNN

Content status: Placeholder.

This subsection has not yet been drafted or technically reviewed.

16.2.2 One-stage detectors: YOLO family, SSD, RetinaNet

Content status: Placeholder.

This subsection has not yet been drafted or technically reviewed.

16.2.3 Anchor-free detectors: FCOS, CenterNet

Content status: Placeholder.

This subsection has not yet been drafted or technically reviewed.

16.2.4 Transformer-based detection: DETR

Content status: Placeholder.

This subsection has not yet been drafted or technically reviewed.

16.2.5 Evaluation: IoU, mAP, per-class analysis

Content status: Placeholder.

This subsection has not yet been drafted or technically reviewed.

16.3 Semantic and instance segmentation

16.3.1 Fully convolutional networks

Content status: Placeholder.

This subsection has not yet been drafted or technically reviewed.

16.3.2 U-Net and encoder–decoder architectures

Content status: Placeholder.

This subsection has not yet been drafted or technically reviewed.

16.3.3 DeepLab and dilated convolutions

Content status: Placeholder.

This subsection has not yet been drafted or technically reviewed.

16.3.4 Mask R-CNN for instance segmentation

Content status: Placeholder.

This subsection has not yet been drafted or technically reviewed.

16.3.5 Panoptic segmentation

Content status: Placeholder.

This subsection has not yet been drafted or technically reviewed.

16.4 Anomaly and defect detection

16.4.1 Reconstruction-based methods: autoencoders and GANs

Content status: Placeholder.

This subsection has not yet been drafted or technically reviewed.

16.4.2 Embedding-based methods: PatchCore, PaDiM, SPADE

Content status: Placeholder.

This subsection has not yet been drafted or technically reviewed.

16.4.3 One-class classifiers and Deep SVDD

Content status: Placeholder.

This subsection has not yet been drafted or technically reviewed.

16.4.4 Benchmarks: MVTec AD and industrial extensions

Content status: Placeholder.

This subsection has not yet been drafted or technically reviewed.

16.4.5 Choosing between anomaly detection and supervised approaches

Content status: Placeholder.

This subsection has not yet been drafted or technically reviewed.