Contrast is an implementation of the generalized contrast function described in:
Reference Paper
(2020, in submission)
A.M.Rodriguez, R.Granger
Dartmouth College
Abstract
- Visual clutter affects our ability to see: objects that
would be identifiable on their own, may become unrecognizable when presented close together (“crowding”) – but the psychophysical characteristics of crowding have resisted simplification. Image properties initially thought to produce crowding have paradoxically yielded unexpected results, e.g., adding flanking objects can ameliorate crowding (Manassi, Sayim et al. 2012, Herzog, Sayim et al. 2015, Pachai, Doerig et al. 2016) The resulting theory revisions have been sufficiently complex and specialized as to make it difficult to discern what principles may underlie the observed phenomena. A generalized formulation of simple visual contrast energy is presented, arising from straightforward analyses of center and surround neurons in the early visual stream. Extant contrast measures, such as RMS contrast, are easily shown to fall out as reduced special cases. The new generalized contrast energy metric surprisingly predicts the principal findings of a broad range of crowding studies. These early crowding phenomena may thus be said to arise predominantly from contrast, or are, at least, severely confounded by contrast effects. (These findings may be distinct from accounts of other, likely downstream, “configural” or “semantic” instances of crowding, suggesting at least two separate forms of crowding that may resist unification.) The new fundamental contrast energy formulation provides a candidate explanatory framework that addresses multiple psychophysical phenomena beyond crowding.
This program is written in Python 3.7.3. https://www.python.org/
- numpy numpy (1.18.1)
- scipy scipy (1.4.1)
- Pillow Pillow (5.4.1) https://pillow.readthedocs.io/en/latest/installation.html
- matplotlib matplotlib (3.1.2)
- pandas pandas (0.25.3)
- PsychoPy Psychopy (2020.1.2)
Install the latest version of these libraries:
$ pip3 install numpy scipy Pillow matplotlib pandas psychopyRunning the Herzog experiment: -------
To run Herzog experiment from the paper:
$ cd experiments/herzog
$ ./herzog.pyResults will be in the report directory:
$ ls report
contrast-Herzog-2012-Figure-1a.pdf
contrast-Herzog-2012-Figure-1b.pdf
contrast-Herzog-2012-Figure-1c.pdf
contrast-Herzog-2012-Figure-1d.pdf
"decision-'herzog'-Figure-1a.pdf"
"decision-'herzog'-Figure-1b.pdf"
"decision-'herzog'-Figure-1c.pdf"
"decision-'herzog'-Figure-1d.pdf"
Herzog-2012-Figure-1a.pdf
Herzog-2012-Figure-1b.pdf
Herzog-2012-Figure-1c.pdf
Herzog-2012-Figure-1d.pdf
$ To run the Herzog experiment with a new est_max value one can edit the parameter directly in the source code or one can issue the following command with a command-line option:
$ cd experiments/herzog
$ ./herzog.py -model est_max=0.1All the parameters are listed at the top of each experiment file. The E_alpha paramer mentioned in the paper appears in the code as est_max. For the Herzog experiment the parameters are in herzog.py:
#foreground_color = [-1,-1,-1]
background_color = [-1,-1,-1]
eccentricity = 3.88
herzog_params = Params(
name = 'herzog',
expName = 'herzog',
viewing_distance = 75.0,
monitor = 'testMonitor',
logfile = 'herzog.log',
window_color = [-1,-1,-1],
exp_info = {u'session': u'001', u'participant': u'default'},
cwd = os.getcwd(),
target_identifier= ('num_flank',0),
stair = Params(startVal=0, stepSizes=1, stepType='lin',
nReversals=0, nTrials=4, nUp=1, nDown=1,
minVal=0, maxVal=7, autoLog=True,
originPath=-1, name='staircase_trials'),
levels = [
Params(exp_num=[1],
num_flank=[0,1,2,4,8],
jitter=[0],
flank_target_height_ratio=[0.5]),
Params(exp_num=[2],
num_flank=[0,1,2,4,8],
jitter=[0],
flank_target_height_ratio=[1]),
Params(exp_num=[3],
num_flank=[0,1,2,4,8],
jitter=[0],
flank_target_height_ratio=[2]),
Params(exp_num=[4],
num_flank=[0,1,2,4,8],
jitter=[1],
flank_target_height_ratio=[0.5])],
experiment = Params(eccentricity= eccentricity,
nTrialReps= 2,
nStaircaseTrials= 8),
stimulus = Params(eccentricity= eccentricity,
jitters=np.array([-0.1, 0.26, -0.87, 0.24,
0.86, -0.34, 0.5 , -0.51])*0.5*(40/60.0),
flank_distance=23.33/60.0,
target_orientation= 0,
line_height= 40/60.0,
line_width= 4/60.0,
vertical_gap= 4/60.0,
offset= 0.0,
filename= ['num_flank','jitter','offset','flank_target_height_ratio','target_orientation'],
offset_level= 16.66/60.0,
offsets= np.array([16.66, 19.04, 21.42, 23.8,
26.18, 28.56, 30.94, 33.32])),
model = Params(eccentricities= [eccentricity], # in deg
view_size= (600,600), # in pixels
view_pos= (eccentricity,0), # center in degrees of visual angle
est_max= 0.1,
upper_limit= 0.85,
lower_limit= 0.0))To recreate the stimuli for Herzog experiment (note: various windows will appear while the stimuli are being generated):
$ cd experiments/herzog
$ ./herzog.py -genstimRunning the Kahneman experiment: -------
To run Herzog experiment from the paper:
$ cd experiments/herzog
$ ./herzog.pyResults will be in the report directory:
$ ls report
contrast-Kahneman-2012-Figure-1.pdf
"decision-'kahneman'-Figure-1.pdf"
Kahneman-2012-Figure-1.pdf
$ All the parameters are listed at the top of each experiment file. For the Kahneman experiment the parameters are in kahneman.py:
foreground_color = [-1,-1,-1]
background_color = [0.1,0.1,0.1]
eccentricity = 0.0
kahneman_params = Params(
name = 'kahneman',
expName = 'kahneman',
exp_num = 1,
viewing_distance = 2300.0,
monitor = 'testMonitor',
logfile = 'kahneman.log',
window_color = background_color,
cwd = os.getcwd(),
exp_info = {u'session': u'001', u'participant': u'default'},
target_identifier= ('flank_distance',-1/60.0),
levels = Params(
flank_distance = np.array([-1., 0.06, 0.12, 0.18, 0.24, 0.6, 1.2, 1.8, 2.4, 3., 5.4 ])/60,
offset = [0],
target_orientation = [0,90,180,270]), # degrees from noon orientation
experiment = Params(eccentricity= eccentricity,
nTrialReps= 1),
stimulus = Params(
eccentricity = eccentricity,
target_size = 0.0548,
gap_size = 0.01124,
line_width= 0.014),
model = Params(eccentricities= [5], # in deg
view_size= (1000,1000), # in pixels
view_pos= (eccentricity,0), # center in degrees of visual angle
est_max= 0.032,
upper_limit= 0.85,
lower_limit= 0.0))To recreate the stimuli for Kahneman experiment (note: various windows will appear while the stimuli are being generated):
$ cd experiments/kahneman
$ ./kahneman.py -genstimRunning the Pachai experiment: -------
To run Pachai experiment from the paper:
$ cd experiments/pachai
$ ./pachai.pyResults will be in the report directory:
$ ls report
contrast-1-Pachai-Figure-1.pdf
contrast-5-Pachai-Figure-1.pdf
"decision-1-'pachai'-Figure-a.pdf"
"decision-5-'pachai'-Figure-a.pdf"
Pachai-1-Figure-1.pdf
Pachai-5-Figure-1.pdf
"plot-'pachai'-barplot.pdf"
$ All the parameters are listed at the top of each experiment file. For the Pachai experiment the parameters are in pachai.py:
eccentricity = 10.0
#background_color = [-1,-1,-1]
background_color = [0,0,0]
pachai_params = Params(
name = 'pachai',
expName = 'pachai',
exp_num = 1,
viewing_distance = 58,
monitor = 'testMonitor',
logfile = 'pachai.log',
exp_info = {u'session': u'001', u'participant': u'default'},
cwd = os.getcwd(),
window_color = background_color,
target_identifier= ('flank_distance',-1),
levels = Params(flank_distance= [-1,0.5,0.9,1.62,2.58,3.9], # degrees of visual angle
flank_orientation= [45,135,225,315], # degrees from noon orientation
target_orientation= [0,90,180,270], # degrees from noon orientation
gap= [0,1],
num_flank= [1,5]),
experiment = Params(eccentricity= eccentricity,
nTrialReps= 1),
stimulus = Params(line_width= 0.4,
gap_width= 1.2,
target_diameter= 2.0,
flank_height= 10.0),
model = Params(eccentricities= [eccentricity], # in deg
view_size= (600,600), # in pixels
view_pos= (eccentricity,0), # center in degrees of visual angle
upper_limit= 0.85,
lower_limit= 0.0))To recreate the stimuli for Pachai experiment (note: various windows will appear while the stimuli are being generated):
$ cd experiments/pachai
$ ./pachai.py -genstimRunning the Pelli experiment: -------
To run Pelli experiment from the paper:
$ cd experiments/pelli
$ ./pelli.pyResults will be in the report directory:
$ ls report
"contrast-'LargeLetters'-Figure-1.pdf"
"contrast-'SmallLetters'-Figure-1.pdf"
"decision-'LargeLetters'-Figure-1.pdf"
"decision-'SmallLetters'-Figure-1.pdf"
"'LargeLetters'-Figure-1.pdf"
"'SmallLetters'-Figure-1.pdf"
$ All the parameters are listed at the top of each experiment file. For the Herzog experiment the parameters are in herzog.py:
background_color = [1,1,1]
eccentricity = [5,10,15,20]
pelli_params = Params(
name = 'pelli',
expName = 'pelli',
exp_num = 1,
viewing_distance = 22*2.54,
monitor = 'testMonitor',
logfile = 'pelli.log',
exp_info = {u'session': u'001', u'participant': u'default'},
cwd = os.getcwd(),
window_color = background_color,
target_identifier= ('flank_distance',-1.0),
levels = Params(
flank_distance = np.array([-1.0,0.05,0.1,0.15,0.2,0.3,0.4,0.6]),
offset = eccentricity),
experiment = Params(
nTrialReps= 2),
stimulus = Params(
name = '',
target_size = 1.0,
gap_size = 0.4,
stim = '02',
line_width= 0.4),
model = Params(eccentricities= eccentricity, # in deg
view_size= (500,500), # in pixels
view_pos= (0,0), # center in degrees of visual angle
upper_limit= 0.85,
lower_limit= 0.0))Since we are using two different decision function in the Pelli case, one for 'Small Letters' and one for the 'LargeLetters', the target_contrast and est_max for those values are defined in the following code fragment at the bottom of the pelli.py file:
if name == 'SmallLetters':
pelli_params['model']['target_contrast'] = 0.005
pelli_params['model']['est_max'] = 0.01
else:
pelli_params['model']['target_contrast'] = 0.012
pelli_params['model']['est_max'] = 0.025 To recreate the stimuli for Pelli experiment (note: various windows will appear while the stimuli are being generated):
$ cd experiments/pelli
$ ./pelli.py -genstim