Open source, data & other downloads

• • Assessment of Physical Rehabilitation Exercises from 3D Skeleton Representations (from Kourbane et al., "SSL-Rehab: Assessment of Physical Rehabilitation Exercises Through Self-Supervised Learning of 3D Skeleton Representations", Computer Vision and Image Understanding, 2025)

• • Image class incremental learning (from Ferdinand et al., "Feature Expansion and enhanced Compression for Class Incremental Learning", Neurocomputing, 2025)

• • CNN ensembles for classification of benthic images (from Vega et al., "Convolutional Neural Networks for Hydrothermal Vents Substratum Classification: an Introspective Study", Ecological Informatics, 2024)

• • Data-driven control of mobile robot manipulators via reinforcement learning (from Mitriakov et al., "An open-source software framework for reinforcement learning-based control of tracked robots in simulated indoor environments", Advanced Robotics, 2022)

• The CAnonically Posed 3D Objects Dataset

The dataset contains a total of 180 generic objects uniformly distributed into 15 classes, wherein objects of the same class share a semantically defined fixed canonical pose. This is guaranteed via the construction protocol of the dataset, as is described in the corresponding article. Objects of the same class differ by body articulations, extrusions, parts deletion/addition, and generally, non-isotropic transformations that alter a specific object part. Transformations that have been applied to generate class instances are applied locally onto the reference object' s shape, therefore despite the fact that a purely geometrically defined canonical pose may is altered, the semantically defined, global canonical pose remains the same. You may refer for more details on the construction of the dataset and its intended use cases in the following article:

Click here to download the dataset.

• • NPCA Rotation normalization (From Papadakis et al.,  "Efficient 3D Shape Matching and Retrieval using a Concrete Radialized Spherical Projection Representation", Pattern Recognition 2007).

• • PANORAMA 3d shape descriptor (From Papadakis et al., "PANORAMA: A 3D Shape Descriptor based on Panoramic Views for Unsupervised 3D Object Retrieval", International Journal of Computer Vision, 2010).

   Suggested use: 

- The motivation in providing the PANORAMA 3D shape descriptor is to allow comparative evaluations with other methods in the domain of unsupervised 3D object retrieval. The descriptor was developed in an effort to make it as discriminative as possible and transparent to any particular 3D objects' dataset or object category. As with any descriptor, PANORAMA could as well be trained and fine-tuned, however, the executable provided here is not intended for such a use. Reverse engineering of the PANORAMA descriptor in order to isolate its features is not recommended, although, not disallowed.

- If you use PANORAMA descriptor, i would appreciate if you contacted me to let me know about your application/project and how PANORAMA allowed you to advance in your work.

- And last, but not least, if you use the PANORAMA descriptor in your work, please consider citing the corresponding journal article as given above.