FN logoFN logo

The Manufacturing Relevant Image Database

20454 images remaining!

Top Leaders

16809

Erick B

9315

Brennan S

7508

James S

1Erick B16809
2Brennan S9315
3James S7508
4William B4350
5Jennifer R3115
6Nathaniel W2410
7Andrew G1845
8Andrew B1760
9Nathan H1671
10James Hardin Hardin Hardin Hardin Hardin Hardin H1667
11Jonathan C1661
12Yash K706
13Andrew F250
14Brandon D126
15Kristiana Mitchell (EXT) 55
16Andy Doyle (EXT) 6
17Kristiana Mitchell (EXT) 6
18Brennan Swick (CTR) 6
19Dan B4
20Nathaniel W0

FactoryNet is a project from the Digital Manufacturing Research Team at the Air Force Research Laboratory (AFRL) that aims to provide a repository of searchable, human labeled, manufacturing relevant images. The most innovative and strategic advancements in manufacturing technology are leveraging digital data to increase efficiency, eliminate errors, and realize agility. A common thread across such efforts is to integrate AI into manufacturing processes and factory environments via new technologies such as augmented reality (AR), virtual reality (VR), and advanced robotics. While AI can be an incredibly powerful tool, it requires reliable and high-quality training data. Public labeled image databases, such as ImageNET and the MNIST handwriting dataset, have become ubiquitous for training and benchmarking algorithms but are often limited for specific applications due to a lack of depth in the associated domains or lack of breadth outside a singular domain. FactoryNET will create a public open image dataset focusing on the manufacturing environment which we believe will accelerate digital manufacturing initiatives at AFRL and across the broader manufacturing community. On this site you can help our mission by labeling images, contributing images to the database and downloading (coming soon) and using our data in your research or practical applications!

We are looking for images of manufacturing relevant objects, preferably in context. When labeling you may be presented with things that don’t fit that description due to the methods of image collection. In these cases, please flag them and we will work to remove them from our dataset. Here are some examples of what we are looking for as relevant images: Machines Hand tools Power tools Construction tools Construction vehicles Tools being used Machines operating Factory floors Machine shops Etc. Here are some examples of images that should be marked as irrelevant: Illustrations Logos Portraits/people Food Passenger Vehicles Non-industrial Buildings Historic artifacts Art 3D renderings Etc. Try and err on the side of relevant but you are always welcome to pass on an image without labeling.

Label to the best of your knowledge! We are looking for labels that are nouns describing the objects in the images not actions/verbs. Try to be accurate but specificity is open ended. For instance, you can label a hammer as a “hand tool” and/or “hammer” and/or “ball-peen hammer” and/or “dewalt ball-peen hammer”. Labels are processed after submission and levels of specificity helps us build the best picture of the contents of our dataset. Feel free to submit multiple labels per image or selection, just remember to separate multiple labels with a comma (, ) delimiter. When labeling we offer a tool to segment images into rectangular sub selections like a crop tool. You can select specific objects in a larger image by drawing a box and labeling the object and then submitting the labels for that selection. If no selection is made labels will be applied to the image as a whole. Thanks for your effort! Any contribution is valuable! If you have any questions or concerns please use our Contact Us page.