Representative datasets for zebrafish imaging in Figure 4 of Abdelfattah et al, Science 2019.

registered.tif 		-> raw imaging data corrected for lateral drifts and hot pixels

swim_data.npy		-> Numpy dictionary with below fields. Refer to "swimdata_explanaion.png" for graphical explanations.
				['ch1']			raw swim signals from the left side
				['ch1_filtered']	processed swim signals from the left side
				['ch2']			raw swim signals from the right side
				['ch2_filtered']	processed swim signals from the right side
				['frame_clock']		synchronization signals between camera trigger and swim signals
				['visual stimulus']	timings of visual stimulus presentation. 1=stopped, 2=moving forward


ROI_xy.npy		-> Numpy dictionary (size=N) of XY coordinates of N active neurons.
			   Each field has 2 x p ndarray (p=number of pixels). 1st and 2nd rows = x and y coordinates of pixels.

raw_timecourse 		-> Numpy ndarray (N x t) of fluorescence timeseries from N active neurons.
                           Weighted sum of pixel values after pixel weight optimization (ref methods)

dff_timecourse 		-> Numpy ndarray (N x t) of normalized fluorescence timeseries from N active neurons.
                           -ΔF/F calculated from above weighted sum of pixel values.

spike_timecourse 	-> Numpy ndarray (N x t) of spike timeseries from N active neurons.
                           Values (0 or 1) represent detected spikes using adaptive template matching

spike_detection_limit 	-> Numpy nudarray (N x 1) of imaging time limit over which we did not use spike and ΔF/F information.
                           Reliability of spike detection decreases with time because fluorescence signals bleach during imaging.
		  	   Therefore we set a criteria on spike shapes (ref methods) and set above time points.