For manufacturing R&D and inspection teams

A research map for vision inspection in manufacturing

7,205 papers on industrial visual inspection and defect detection, 2016–2026: wafers and chips, PCBs, displays, batteries, metal surfaces, welds, castings and printed parts. Collected from 26 manufacturing, measurement and vision journals and conferences, plus the anomaly-detection work of the top AI venues.

Bring us your inspection problem How it works
Wafer and chip defectsWeld inspectionMetal surface defectsPCB and solderAdditive manufacturing Your problem lands here
Each dot is a paper; papers on similar problems sit together. (Illustration.)

7,205

papers on manufacturing vision

26 + 10

manufacturing venues + AI venues

13

inspection fields

2016–2026

updated as new papers appear

Fields

Papers are grouped by what is inspected. Semiconductor inspection comes first: wafer maps, die and package defects, SEM review images.

First priority

Semiconductor and wafer

Wafer maps, die, package and bump defects, lithography

433 papers

PCB and electronics

Boards, solder joints, components

206 papers

Displays

LCD and OLED panels, mura

37 papers

Batteries and solar cells

Electrodes, cells, electroluminescence images

166 papers

Metal surfaces

Steel strip, aluminium, pipes, rails

422 papers

Welding

Seams, weld pools, radiographs

900 papers

Casting and forging

Porosity, shrinkage, X-ray and CT

141 papers

Additive manufacturing

Melt pools, powder beds, printed layers

630 papers

Machining and tools

Tool wear, workpiece surfaces, roughness

246 papers

Textiles, composites and other materials

Fabric, CFRP, wood, ceramics, glass

598 papers

Assembly and products

Parts, gears, packaging, labels

495 papers

Anomaly detection in general

MVTec- and VisA-style methods, zero- and few-shot

2,658 papers

Other industrial vision

Cameras, measurement and setups not listed above

273 papers

How it works

1

Describe the problem

What is inspected, which defects, how it is imaged and how many defect samples you have. For example: “scratches and particles on 300 mm wafers, brightfield images, few labelled defects”.

2

See the closest work

scime places the problem on the map and lists the papers nearest to it, across manufacturing journals and AI conferences, which are rarely read together.

3

Choose what to try

Which models, datasets and metrics those papers use, and what they report when they changed one thing. Results are read from open-access full texts.

Coverage

Papers from 2016 to 2026 whose title or abstract is about imaging and about defects or inspection, kept only when they are about manufacturing. Road, bridge and building cracks, medical images, crops and signal-only fault diagnosis are left out.

Manufacturing, measurement and vision venues (26)

IEEE Transactions on Industrial Informatics · Instrumentation and Measurement · Semiconductor Manufacturing · Automation Science and Engineering · Industrial Electronics · Image Processing · IEEE Sensors Journal · Sensors · Measurement · Int. J. of Advanced Manufacturing Technology · Journal of Manufacturing Processes · Journal of Manufacturing Systems · Journal of Intelligent Manufacturing · Journal of Manufacturing Science and Engineering · CIRP Annals · Robotics and Computer-Integrated Manufacturing · Computers in Industry · Advanced Engineering Informatics · Engineering Applications of AI · Expert Systems with Applications · NDT & E International · Mechanical Systems and Signal Processing · CVPR · WACV · ICIP · IEEE CASE

From the scime AI Paper Atlas (10)

Industrial anomaly detection and inspection papers from CVPR · ICCV · ECCV · NeurIPS · ICML · ICLR · AAAI · TPAMI · IJCV · JMLR

Paper data from OpenAlex (CC0). Abstracts and full texts are used only where their licence allows.

Working on a vision inspection model?

The paper database is used by the scime team. Send us your inspection problem and we will come back with the closest papers, the methods they tried and what they reported.

contact@latenticlabs.com