Heatmap with datagaze

BearGrillz

Hi guys, im having problem with the code for generating heatmap

def draw_heatmap(fixations, dispsize, imagefile=None, durationweight=True, alpha=0.5, savefilename=None):
# FIXATIONS
		fix = parse_fixations(fixations)
		
		# IMAGE
		fig, ax = draw_display(dispsize, imagefile=imagefile)
	

		# HEATMAP
		# Gaussian
		gwh = 200
		gsdwh = gwh/6
		gaus = gaussian(gwh,gsdwh)
		# matrix of zeroes
		strt = gwh/2
		heatmapsize = dispsize[1] + 2*strt, dispsize[0] + 2*strt
		heatmap = numpy.zeros(heatmapsize, dtype=float)
		# create heatmap
		for i in range(0,len(fix['dur'])):
			# get x and y coordinates
			#x and y - indexes of heatmap array. must be integers
			x = strt + int(fix['x'][i]) - int(gwh/2)
			y = strt + int(fix['y'][i]) - int(gwh/2)
			# correct Gaussian size if either coordinate falls outside of
			# display boundaries
			if (not 0 < x < dispsize[0]) or (not 0 < y < dispsize[1]):
				hadj=[0,gwh];vadj=[0,gwh]
				if 0 > x:
					hadj[0] = abs(x)
					x = 0
				elif dispsize[0] < x:
					hadj[1] = gwh - int(x-dispsize[0])
				if 0 > y:
					vadj[0] = abs(y)
					y = 0
				elif dispsize[1] < y:
					vadj[1] = gwh - int(y-dispsize[1])
				# add adjusted Gaussian to the current heatmap
				try:
					heatmap[y:y+vadj[1],x:x+hadj[1]] += gaus[vadj[0]:vadj[1],hadj[0]:hadj[1]] * fix['dur'][i]
				except:
					# fixation was probably outside of display
					pass
			else:				
				# add Gaussian to the current heatmap
				heatmap[y:y+gwh,x:x+gwh] += gaus * fix['dur'][i]
		# resize heatmap
		heatmap = heatmap[strt:dispsize[1]+strt,strt:dispsize[0]+strt]
		# remove zeros
		lowbound = numpy.mean(heatmap[heatmap>0])
		heatmap[heatmap<lowbound] = numpy.NaN
		# draw heatmap on top of image
		ax.imshow(heatmap, cmap='jet', alpha=alpha)
	

		# FINISH PLOT
		# invert the y axis, as (0,0) is top left on a display
		ax.invert_yaxis()
		# save the figure if a file name was provided
		if savefilename != None:
			fig.savefig(savefilename)
		
		return fig

I don't understand variables 'gwh' , 'gsdwh', 'strt' , 'hadj', 'vadj' stand for. Why we need to use them

Thanks in advance

eduard

Hi,

Can you share the link to the script that uses that function? Then I can try to figure out what they do based on the context. My guess is that gwh and gwsdh are needed to define a Gaussian kernel. strt is needed to extract the gaussian in one dimension, and hadj and vadj stand for horizontal and vertical adjustment, and are mark the border of the Gaussian in pixel coordinates.

But as I said, if you share the script, I can try to give more detailed answeres.

Eduard