Welcome ! And congratulations you made it here.
But do you know why you are here ? Did you come knowingly, or did you just find an open door and thought it seemed nice to come in ? Whichever it is, we are glad to welcome you, and happy to tell you what we do here.
- For hundred of purposes, data is the key:
- machine learning
- algorithm benchmark
- unit testing
- functional testing
- processing pipe design
- algorithm optimization / parameters tweak
- sensor caracterization
- and probably many others...
For all those purposes, having data representing a situation, a problem, a scenario the program must be able to handle is crucial to the achievement of the task. And most of the times, a few data is simply not enough.
qidata
is a set of tools made to handle annotated files and data sets. It helps the world to store
information about recorded data in order to be used as ground truth during testing, or as selection criteria
for machine learning or performance statistics.
It is probably not that much different. However, most of those tools were designed with a very particular area
in mind (machine learning for face detection or OCR, set of recorded speeches for speech recognition, music
database for music classification). We on the other hand, tried our best to be as generic as possible, as our key
goal is to work with robots, which could encounter many different situations and need many different algorithms.
As a result, qidata
can work with images and video but also with lasers, sonars or audio. And most important,
qidata
is thought to handle multi-modal data, whereas most annotation tools available cannot handle more than
one data type.
Find the complete documentation here