RoNIN Dataset
-------------
The dataset contains 327 sequences in total in the format
	sequence_name
		|--- data.hdf5
		|--- info.json

Unfortunately, due to security concerns we were unable to publish ~ 50% of our dataset.


HDF5 data format
-----------------
data.hdf5
     |---raw
     |    |---tango
     |          |---gyro, gyro_uncalib, acce, magnet, game_rv, gravity, linacce, step, tango_pose, tango_adf_pose, rv, pressure, (optional) [wifi, gps, magnetic_rv, magnet_uncalib]
     |    |--- imu
     |          |---gyro, gyro_uncalib, acce, magnet, game_rv, gravity, linacce, step. rv, pressure, (optional) [wifi, gps, magnetic_rv, magnet_uncalib]
     |--synced
     |    |--- time, gyro, gyro_uncalib, acce, magnet, game_rv, rv, gravity, linacce, step
     |---pose
     |    |---tango_pos, tango_ori, (optional) ekf_ori



HDF5 data description
---------------------
Data sources
	 - gyro		 		- Android Sensor.TYPE_GYROSCOPE
	 - gyro_uncalib		- Android Sensor.TYPE_GYROSCOPE_UNCALIBRATED
	 - acce		 		- Android Sensor.TYPE_ACCELEROMETER
	 - linacce			- Android Sensor.TYPE_LINEAR_ACCELERATION
	 - gravity			- Android Sensor.TYPE_GRAVITY
	 - magnet			- Android Sensor.TYPE_MAGNETIC_FIELD
	 - magnet_uncalib 	- Android Sensor.TYPE_MAGNETIC_FIELD_UNCALIBRATED
	 - rv		 		- Android Sensor.TYPE_ROTATION_VECTOR
	 - game_rv			- Android Sensor.TYPE_GAME_ROTATION_VECTOR
	 - magnetic_rv		- Android Sensor.TYPE_GEOMAGNETIC_ROTATION_VECTOR
	 - step		 		- Android Sensor.TYPE_STEP_COUNTER
	 - pressure			- Android Sensor.TYPE_PRESSURE
	 - gps				- Android LocationManager.GPS_PROVIDER
	 - tango_pose		- Pose from Visual SLAM of Tango device (format: time, position (x,y,z), orientation (x,y,z,w))
	 - tango_adf__pose	- Pose from Visual SLAM with area learning of Tango device
	 - wifi				- wifi footprints scanned every 3 seconds. stored in 2 parts
		|-- "wifi_values" - contains (scan_number, last_timestep, level) 
		|--"wifi_address" - dataset of type string. contains BSSID of corresponding records in wifi_values

"raw" group contains data as reported by APIs in format 
	- system_timestamp (nanosecond), API output
"synced" group contains time synchronized data from IMU device sampled at 200 Hz
   	- time : System time of IMU device in seconds
"pose" group store all pose information (timestamp for data is "synced/time")
   	- tango_pos, tango_ori - tango_pose and tango_adf_pose combined for more accurate pose estimation
   	- ekf_ori - orientation of IMU device calculated using sensor fusion. (Should not be used during testing)


JSON data description
---------------------
Stores meta information
	 - length	: sequence length in seconds
	 - date  	: collected date mm/dd/yy
	 - device	: device_identifier

	 - align_tango_to_body : approx. alignment between Tango device coordinate frame and body 
	 - start_frame		   : dataframe which marks end of pre-calibration and synchronization. (use data from this point forward)

	Sensor calibration of IMU device. (Please refer supplementary material for details)	
	 - imu_init_gyro_bias
	 - imu_end_gyro_bias 
	 - imu_acce_bias
	 - imu_acce_scale

	Time syncronization between the system time of two devices 
	t_imu = t_tango - tango_reference_time  + imu_reference_time + imu_time_offset
	 - imu_reference_time	: approx. time alignments (nanoseconds) 
	 - tango_reference_time	: approx. time alignments (nanoseconds) 
	 - imu_time_offset: 	: fine tuned time alignments (seconds)

	Alignment between Tango refernce coordinate frame and game rotation vector of IMU device
	 - start_calibration 
	 - end_calibration

	Orientation drift caluculated using end_calibration 
	 - gyro_integration_error
	 - grv_ori_error		 
	 - ekf_ori_error