I'm looking for a dataset that would be the basis for benchmarking pose detection algorithms (and even dead-reckoning positioning algorithms).
Some datasets I find are just 3 or 6 DOF (accelerometers+gyroscopes).
Others do no provide a ground truth reference (an accurate reference of the position and orientation).
My ideal dataset would consist on a set of sensor measurements (9 DOF), and the corresponding position vector and orientation info (either Euler, or quaternions) along time (with some given sampling freq).
So, a simple CSV table, with the following header:
AX, AY, AZ, GX, GY, GZ, MX, MY, MZ, X, Y, Z, ROLL, PITCH, YAW
Maybe the table should be split in different ones due to the difference in sampling frequencies of the sensors, and probably the ability to get ground truth readings with the same resolution (if they are acquired by a camera).
Hello David. If you are still looking for a dataset, you might want to check this ones.
1) First one: http://gaitanalysis.th-brandenburg.de/
From that website you can also access the paper associated with it: http://gaitanalysis.th-brandenburg.de/static/files/GaitAnalysisDataBaseShortOverview.f883b62f9e67.pdf
2) Second one: http://igt.aei.polsl.pl/index.php/en/research/projects/61-repoimu
And this is the paper: Conference Paper Reference Data Set for Accuracy Evaluation of Orientation Es...
3) This third one is not what you are looking for, but maybe it is useful for something else: Article Benchmark Datasets for Bilateral Lower-Limb Neuromechanical ...
Hope it's helpful.
The answer from Fernando Kabas is very good, it already has some pointers. On top of that, we are actually building our own (robot based) dataset which we want to make public.
- Badges
- Science topic
- Similar topics
- Geoscience
- Remote Sensing