Welcome to the Sparse Constellations Dataset and Code Repository. Here you'll find an example API to access the dataset using python and numpy.
This dataset contains about 640K annotated instances, segmented by day, room, and location. Each data instance contains roughly 1207 features. It also contains data for in-room events, as well as synchronized data streams from all other sensors.
python3, numpy, and pickle
The dataset files are hosted on dropbox (too big for Github). Download them here: https://www.dropbox.com/sh/bxx86a79we730ic/AAD1ENqMMrKgAsXBXkgF-d-Ra?dl=0
The data folder contains python serialized classes of all collected data. Make sure you download the files above, and save it as '/data'.
sample_loop.py lists the neccessary routines for acessing the data.
The basic steps for accessing the raw numpy arrays are as follows:
- Unpickle the data file for a room using:
room_dataset = load_data('data.pklz')- To access data instances for that room, use:
room_dataset['Day1']['data'] for 'Day1' data. Data goes up to 'Day7'- To access data labels, use:
data = room_dataset['Day1']['labels']