LOCO: Logistics Objects in Context

Deep Learning systems are on the rise, challenging state-of-the-art solutions in almost all areas of application, including computer vision, natural language processing, playing atari games and folding proteins. While unsupervised learning mechanisms are being researched, supervised learning is still the predominant learning strategy, especially when considering application-oriented systems. In contrast to unsupervised learning, supervised learning demands labelled data in order to learn given patterns. Since the rise of data,

Depending on the type of data, capturing and labelling data is quite a tedious task.

While the amount of data necessary for training a sufficient model has definitely decreased due to e.g. transfer learning and the broad availability of pretrained models, training data is still a major component of current learning systems.

LOCO is an image dataset targeting object detection in logistics scenes. Images are captured while walking through a logistics setting using low-cost consumer cameras.

LOCO currently provides XXX total images captured in 5 logistics environments, of which XXXX images are manually annotated. Annotated classes include pallets, small load carriers, pallet trucks, forklifts and stillages.

For more details, we refer the reader to the official paper. If you use LOCO for your work, consider citeing (see bibtex below).