brainscore-vision 2.1__py3-none-any.whl

brainscore_vision/data/geirhos2021/data_packaging/edge/edge_stimulus_set.py

@@ -0,0 +1,69 @@
+from pathlib import Path
+from brainio.stimuli import StimulusSet
+from brainio.packaging import package_stimulus_set
+
+stimuli = []
+image_paths = {}
+stimuli_directory = '../datasets/edge'
+
+
+'''
+Dataset Meta Info (from https://github.com/rgeirhos/generalisation-humans-DNNs)
+
+** NOTE: This benchmark (edge) has a very different stimulus sets then others!! **
+
+Sample image from dataset:
+bear4.png
+
+This is a concatenation of the following information:
+
+    1) Ground Truth category in the 16-set of:
+        {airplane, bear, bicycle, bird, boat, bottle, car, cat, 
+        chair, clock, dog, elephant, keyboard, knife, oven, truck} 
+    2) The sub-image number, in the set {1, 2, 3, ... 10}
+    
+There are thus 100 total images in this stimulus set (16 categories * 10 variations = 160) 
+
+'''
+
+for filepath in Path(stimuli_directory).glob('*.png'):
+
+    # entire name of image file:
+    image_id = filepath.stem
+
+    # re to parse image names into component parts:
+    import re
+    match = re.match(r"([a-z]+)([0-9]+)", image_id, re.I)
+    if match:
+        items = match.groups()
+    else:
+        items = []
+
+    split_name = items
+
+    # ensure proper metadata length per image in set
+    assert len(split_name) == 2
+
+    # Dataset image data, 1-7 from above:
+    image_category = split_name[0]
+    image_variation = split_name[1]
+
+    image_paths[image_id] = filepath
+    stimuli.append({
+        'image_id': image_id,
+        'image_category': image_category,
+        'truth': image_category,
+        'image_variation': image_variation,
+        'condition': 0,
+    })
+
+stimuli = StimulusSet(stimuli)
+stimuli.image_paths = image_paths
+stimuli.name = 'Geirhos2021_edge'  # give the StimulusSet an identifier name
+
+# Ensure 100 images in dataset
+assert len(stimuli) == 160
+
+# upload to S3
+package_stimulus_set("brainio_brainscore", stimuli, stimulus_set_identifier=stimuli.name,
+                     bucket_name="brainio-brainscore")

brainscore_vision/data/geirhos2021/data_packaging/eidolonI/eidolonI_data_assembly.py

@@ -0,0 +1,92 @@
+import numpy as np
+
+from brainio.assemblies import BehavioralAssembly
+from brainio.packaging import package_data_assembly
+import pandas as pd
+
+'''
+Experiment Information:
+https://arxiv.org/pdf/1706.06969.pdf
+
+ - 4 subjects 
+ - 1280 images each
+ - 5120 total images shown 
+ - match to sample task, 16AFC
+ - 16 image categories
+ - for the this benchmark (eidolonI) subjects saw the EXACT image indicated with the variable/column name
+   image_lookup_id, and not a variation of it (no distortions, editing, etc). Condition is a three-number
+   set, like 129-10-10, describing the eidolonI parameters:
+   
+   eidolon-experiment: in the form 'a-b-c', indicating:
+
+   a is the parameter value for 'reach', in {1,2,4,8,...128}
+   b in {0,3,10} for coherence value of 0.0, 0.3, or 1.0; these are referred 
+   to as Eidolon III, Eidolon II and Eidolon I in the paper.
+   c = 10 for grain value of 10.0 (not varied in this experiment)
+
+'''
+
+# initial csv to dataframe processing:
+subject_1 = pd.read_csv('data_assemblies/eidolonI_subject-01_session_1.csv')
+subject_2 = pd.read_csv('data_assemblies/eidolonI_subject-02_session_1.csv')
+subject_3 = pd.read_csv('data_assemblies/eidolonI_subject-03_session_1.csv')
+subject_4 = pd.read_csv('data_assemblies/eidolonI_subject-04_session_1.csv')
+
+all_subjects = pd.concat([subject_1, subject_2, subject_3, subject_4])
+
+# parse df for the image lookup id. This relates the data assembly with the stimulus set.
+split_cols = all_subjects['imagename'].str.split("_", expand=True)
+drop_cols = split_cols.drop(split_cols.columns[[0, 1, 2]], axis=1)
+all_subjects['image_lookup_id'] = drop_cols.agg("_".join, axis=1).str.replace(".png", "")
+
+
+# construct the assembly
+assembly = BehavioralAssembly(all_subjects['object_response'],
+                              coords={
+                                  'image_id': ('presentation', all_subjects['image_lookup_id']),
+                                  'image_id_long': ('presentation', all_subjects['imagename']),
+                                  'truth': ('presentation', all_subjects['category']),
+                                  'choice': ('presentation', all_subjects['object_response']),
+                                  'category': ('presentation', all_subjects['category']),
+                                  'condition': ('presentation', all_subjects['condition']),
+                                  'response_time': ('presentation', all_subjects['rt']),
+                                  'trial': ('presentation', all_subjects['trial']),
+                                  'subject': ('presentation', all_subjects['subj']),
+                                  'session': ('presentation', all_subjects['Session']),
+                              },
+                              dims=['presentation']
+                              )
+
+# give the assembly an identifier name
+assembly.name = 'brendel.Geirhos2021_eidolonI'
+
+# make sure assembly dim is correct length
+assert len(assembly['presentation']) == 5120
+
+# make sure assembly coords are correct length
+assert len(assembly['image_id']) == 5120
+assert len(assembly['image_id_long']) == 5120
+assert len(assembly['truth']) == 5120
+assert len(assembly['category']) == 5120
+assert len(assembly['condition']) == 5120
+assert len(assembly['response_time']) == 5120
+assert len(assembly['trial']) == 5120
+assert len(assembly['subject']) == 5120
+assert len(assembly['session']) == 5120
+
+
+# make sure there are 1280 unique images (shown 1 time for each  of 4 subjects, total of 4 * 1280 = 5120 images shown)
+assert len(np.unique(assembly['image_id'].values)) == 1280
+
+# make sure there are 4 unique subjects:
+assert len(np.unique(assembly['subject'].values)) == 4
+
+# make sure there are 16 unique object categories (ground truths):
+assert len(np.unique(assembly['truth'].values)) == 16
+assert len(np.unique(assembly['category'].values)) == 16
+
+
+# upload to S3
+package_data_assembly('brainio_brainscore', assembly, assembly_identifier=assembly.name,
+                      stimulus_set_identifier="Geirhos2021_eidolonI",
+                      assembly_class="BehavioralAssembly", bucket_name="brainio-brainscore")

brainscore_vision/data/geirhos2021/data_packaging/eidolonI/eidolonI_stimulus_set.py

@@ -0,0 +1,82 @@
+from pathlib import Path
+from brainio.stimuli import StimulusSet
+from brainio.packaging import package_stimulus_set
+
+stimuli = []
+image_paths = {}
+stimuli_directory = '../datasets/eidolonI/dnn/session-1'
+
+
+'''
+Dataset Meta Info (from https://github.com/rgeirhos/generalisation-humans-DNNs)
+
+Sample image from dataset:
+0001_eid_dnn_1-10-10_oven_10_n04111531_17072.png
+
+This is a concatenation of the following information (separated by '_'):
+
+    1) a four-digit number starting with 0000 for the first image in an experiment; 
+       the last image therefore has the number n-1 if n is the number of images in a certain experiment
+    2) short code for experiment name, e.g. 'eid' for eidolon-experiment
+    3) either e.g. 's01' for 'subject-01', or 'dnn' for DNNs
+    4) condition
+    5) category (ground truth)
+    6) a number (just ignore it)
+    7) image identifier in the form a_b.JPEG (or a_b.png), with a being the 
+       WNID (WordNet ID) of the corresponding synset and b being an integer.
+'''
+
+for filepath in Path(stimuli_directory).glob('*.png'):
+
+    # entire name of image file:
+    image_id = filepath.stem
+    image_id_long = image_id
+    split_name = filepath.stem.split('_')
+
+    # ensure proper metadata length per image in set
+    assert len(split_name) == 8
+
+    # Dataset image data, 1-7 from above:
+    image_number = split_name[0]
+    experiment_code = split_name[1]
+    subject = split_name[2]
+    condition = split_name[3]
+    category_ground_truth = split_name[4]
+    random_number = split_name[5]
+
+    # note the split, for a total of 8 metadata fields:
+    wordnet_a = split_name[6]
+    wordnet_b = split_name[7]
+
+    # image lookup ID, same as data assembly. This is the exact image shown to participant.
+    # This is needed, as the raw images have "dnn" in the subject field, even when a human subject
+    # was used. Otherwise, the image name data in the raw data table and the image names themselves are the same.
+    image_lookup_id = "_".join(split_name[3:])
+
+    image_paths[image_id] = filepath
+    stimuli.append({
+        'image_id': image_lookup_id,
+        'image_id_long': image_id_long,
+        'image_number': image_number,
+        'experiment_code': experiment_code,
+        'condition': condition,
+        'truth': category_ground_truth,
+        'category_ground_truth': category_ground_truth,
+        'random_number': random_number,
+        'wordnet_a': wordnet_a,
+        'wordnet_b': wordnet_b,
+    })
+
+stimuli = StimulusSet(stimuli)
+image_id_to_lookup = dict(zip(stimuli['image_id_long'], stimuli['image_id']))
+stimuli.image_paths = image_paths
+stimuli.image_paths = {image_id_to_lookup[image_id]: path
+                       for image_id, path in stimuli.image_paths.items()}
+stimuli.name = 'Geirhos2021_eidolonI'  # give the StimulusSet an identifier name
+
+# Ensure 1280 images in dataset
+assert len(stimuli) == 1280
+
+# upload to S3
+package_stimulus_set("brainio_brainscore", stimuli, stimulus_set_identifier=stimuli.name,
+                     bucket_name="brainio-brainscore")

brainscore_vision/data/geirhos2021/data_packaging/eidolonII/eidolonII_data_assembly.py

@@ -0,0 +1,92 @@
+import numpy as np
+
+from brainio.assemblies import BehavioralAssembly
+from brainio.packaging import package_data_assembly
+import pandas as pd
+
+'''
+Experiment Information:
+https://arxiv.org/pdf/1706.06969.pdf
+
+ - 4 subjects 
+ - 1280 images each
+ - 5120 total images shown 
+ - match to sample task, 16AFC
+ - 16 image categories
+ - for the this benchmark (eidolonII) subjects saw the EXACT image indicated with the variable/column name
+   image_lookup_id, and not a variation of it (no distortions, editing, etc). Condition is a three-number
+   set, like 129-10-10, describing the eidolonI parameters:
+   
+   eidolon-experiment: in the form 'a-b-c', indicating:
+
+   a is the parameter value for 'reach', in {1,2,4,8,...128}
+   b in {0,3,10} for coherence value of 0.0, 0.3, or 1.0; these are referred 
+   to as Eidolon III, Eidolon II and Eidolon I in the paper.
+   c = 10 for grain value of 10.0 (not varied in this experiment)
+
+'''
+
+# initial csv to dataframe processing:
+subject_1 = pd.read_csv('data_assemblies/eidolonII_subject-01_session_1.csv')
+subject_2 = pd.read_csv('data_assemblies/eidolonII_subject-02_session_1.csv')
+subject_3 = pd.read_csv('data_assemblies/eidolonII_subject-03_session_1.csv')
+subject_4 = pd.read_csv('data_assemblies/eidolonII_subject-04_session_1.csv')
+
+all_subjects = pd.concat([subject_1, subject_2, subject_3, subject_4])
+
+# parse df for the image lookup id. This relates the data assembly with the stimulus set.
+split_cols = all_subjects['imagename'].str.split("_", expand=True)
+drop_cols = split_cols.drop(split_cols.columns[[0, 1, 2]], axis=1)
+all_subjects['image_lookup_id'] = drop_cols.agg("_".join, axis=1).str.replace(".png", "")
+
+
+# construct the assembly
+assembly = BehavioralAssembly(all_subjects['object_response'],
+                              coords={
+                                  'image_id': ('presentation', all_subjects['image_lookup_id']),
+                                  'image_id_long': ('presentation', all_subjects['imagename']),
+                                  'truth': ('presentation', all_subjects['category']),
+                                  'choice': ('presentation', all_subjects['object_response']),
+                                  'category': ('presentation', all_subjects['category']),
+                                  'condition': ('presentation', all_subjects['condition']),
+                                  'response_time': ('presentation', all_subjects['rt']),
+                                  'trial': ('presentation', all_subjects['trial']),
+                                  'subject': ('presentation', all_subjects['subj']),
+                                  'session': ('presentation', all_subjects['Session']),
+                              },
+                              dims=['presentation']
+                              )
+
+# give the assembly an identifier name
+assembly.name = 'brendel.Geirhos2021_eidolonII'
+
+# make sure assembly dim is correct length
+assert len(assembly['presentation']) == 5120
+
+# make sure assembly coords are correct length
+assert len(assembly['image_id']) == 5120
+assert len(assembly['image_id_long']) == 5120
+assert len(assembly['truth']) == 5120
+assert len(assembly['category']) == 5120
+assert len(assembly['condition']) == 5120
+assert len(assembly['response_time']) == 5120
+assert len(assembly['trial']) == 5120
+assert len(assembly['subject']) == 5120
+assert len(assembly['session']) == 5120
+
+
+# # make sure there are 1280 unique images (shown 1 time for each  of 4 subjects, total of 4 * 1280 = 5120 images shown)
+assert len(np.unique(assembly['image_id'].values)) == 1280
+
+# make sure there are 4 unique subjects:
+assert len(np.unique(assembly['subject'].values)) == 4
+
+# make sure there are 16 unique object categories (ground truths):
+assert len(np.unique(assembly['truth'].values)) == 16
+assert len(np.unique(assembly['category'].values)) == 16
+
+
+# upload to S3
+package_data_assembly('brainio_brainscore', assembly, assembly_identifier=assembly.name,
+                      stimulus_set_identifier="Geirhos2021_eidolonII",
+                      assembly_class="BehavioralAssembly", bucket_name="brainio-brainscore")

brainscore_vision/data/geirhos2021/data_packaging/eidolonII/eidolonII_stimulus_set.py

@@ -0,0 +1,82 @@
+from pathlib import Path
+from brainio.stimuli import StimulusSet
+from brainio.packaging import package_stimulus_set
+
+stimuli = []
+image_paths = {}
+stimuli_directory = '../datasets/eidolonII/dnn/session-1'
+
+
+'''
+Dataset Meta Info (from https://github.com/rgeirhos/generalisation-humans-DNNs)
+
+Sample image from dataset:
+0001_eid_dnn_1-10-10_oven_10_n04111531_17072.png
+
+This is a concatenation of the following information (separated by '_'):
+
+    1) a four-digit number starting with 0000 for the first image in an experiment; 
+       the last image therefore has the number n-1 if n is the number of images in a certain experiment
+    2) short code for experiment name, e.g. 'eid' for eidolon-experiment
+    3) either e.g. 's01' for 'subject-01', or 'dnn' for DNNs
+    4) condition
+    5) category (ground truth)
+    6) a number (just ignore it)
+    7) image identifier in the form a_b.JPEG (or a_b.png), with a being the 
+       WNID (WordNet ID) of the corresponding synset and b being an integer.
+'''
+
+for filepath in Path(stimuli_directory).glob('*.png'):
+
+    # entire name of image file:
+    image_id = filepath.stem
+    image_id_long = image_id
+    split_name = filepath.stem.split('_')
+
+    # ensure proper metadata length per image in set
+    assert len(split_name) == 8
+
+    # Dataset image data, 1-7 from above:
+    image_number = split_name[0]
+    experiment_code = split_name[1]
+    subject = split_name[2]
+    condition = split_name[3]
+    category_ground_truth = split_name[4]
+    random_number = split_name[5]
+
+    # note the split, for a total of 8 metadata fields:
+    wordnet_a = split_name[6]
+    wordnet_b = split_name[7]
+
+    # image lookup ID, same as data assembly. This is the exact image shown to participant.
+    # This is needed, as the raw images have "dnn" in the subject field, even when a human subject
+    # was used. Otherwise, the image name data in the raw data table and the image names themselves are the same.
+    image_lookup_id = "_".join(split_name[3:])
+
+    image_paths[image_id] = filepath
+    stimuli.append({
+        'image_id': image_lookup_id,
+        'image_id_long': image_id_long,
+        'image_number': image_number,
+        'experiment_code': experiment_code,
+        'condition': condition,
+        'truth': category_ground_truth,
+        'category_ground_truth': category_ground_truth,
+        'random_number': random_number,
+        'wordnet_a': wordnet_a,
+        'wordnet_b': wordnet_b,
+    })
+
+stimuli = StimulusSet(stimuli)
+image_id_to_lookup = dict(zip(stimuli['image_id_long'], stimuli['image_id']))
+stimuli.image_paths = image_paths
+stimuli.image_paths = {image_id_to_lookup[image_id]: path
+                       for image_id, path in stimuli.image_paths.items()}
+stimuli.name = 'Geirhos2021_eidolonII'  # give the StimulusSet an identifier name
+
+# Ensure 1280 images in dataset
+assert len(stimuli) == 1280
+
+# upload to S3
+package_stimulus_set("brainio_brainscore", stimuli, stimulus_set_identifier=stimuli.name,
+                     bucket_name="brainio-brainscore")

brainscore_vision/data/geirhos2021/data_packaging/eidolonIII/eidolonIII_data_assembly.py

@@ -0,0 +1,92 @@
+import numpy as np
+
+from brainio.assemblies import BehavioralAssembly
+from brainio.packaging import package_data_assembly
+import pandas as pd
+
+'''
+Experiment Information:
+https://arxiv.org/pdf/1706.06969.pdf
+
+ - 4 subjects 
+ - 1280 images each
+ - 5120 total images shown 
+ - match to sample task, 16AFC
+ - 16 image categories
+ - for the this benchmark (eidolonIII) subjects saw the EXACT image indicated with the variable/column name
+   image_lookup_id, and not a variation of it (no distortions, editing, etc). Condition is a three-number
+   set, like 129-10-10, describing the eidolonI parameters:
+   
+   eidolon-experiment: in the form 'a-b-c', indicating:
+
+   a is the parameter value for 'reach', in {1,2,4,8,...128}
+   b in {0,3,10} for coherence value of 0.0, 0.3, or 1.0; these are referred 
+   to as Eidolon III, Eidolon II and Eidolon I in the paper.
+   c = 10 for grain value of 10.0 (not varied in this experiment)
+
+'''
+
+# initial csv to dataframe processing:
+subject_1 = pd.read_csv('data_assemblies/eidolonIII_subject-01_session_1.csv')
+subject_2 = pd.read_csv('data_assemblies/eidolonIII_subject-02_session_1.csv')
+subject_3 = pd.read_csv('data_assemblies/eidolonIII_subject-03_session_1.csv')
+subject_4 = pd.read_csv('data_assemblies/eidolonIII_subject-04_session_1.csv')
+
+all_subjects = pd.concat([subject_1, subject_2, subject_3, subject_4])
+
+# parse df for the image lookup id. This relates the data assembly with the stimulus set.
+split_cols = all_subjects['imagename'].str.split("_", expand=True)
+drop_cols = split_cols.drop(split_cols.columns[[0, 1, 2]], axis=1)
+all_subjects['image_lookup_id'] = drop_cols.agg("_".join, axis=1).str.replace(".png", "")
+
+
+# construct the assembly
+assembly = BehavioralAssembly(all_subjects['object_response'],
+                              coords={
+                                  'image_id': ('presentation', all_subjects['image_lookup_id']),
+                                  'image_id_long': ('presentation', all_subjects['imagename']),
+                                  'truth': ('presentation', all_subjects['category']),
+                                  'choice': ('presentation', all_subjects['object_response']),
+                                  'category': ('presentation', all_subjects['category']),
+                                  'condition': ('presentation', all_subjects['condition']),
+                                  'response_time': ('presentation', all_subjects['rt']),
+                                  'trial': ('presentation', all_subjects['trial']),
+                                  'subject': ('presentation', all_subjects['subj']),
+                                  'session': ('presentation', all_subjects['Session']),
+                              },
+                              dims=['presentation']
+                              )
+
+# give the assembly an identifier name
+assembly.name = 'brendel.Geirhos2021_eidolonIII'
+
+# make sure assembly dim is correct length
+assert len(assembly['presentation']) == 5120
+
+# make sure assembly coords are correct length
+assert len(assembly['image_id']) == 5120
+assert len(assembly['image_id_long']) == 5120
+assert len(assembly['truth']) == 5120
+assert len(assembly['category']) == 5120
+assert len(assembly['condition']) == 5120
+assert len(assembly['response_time']) == 5120
+assert len(assembly['trial']) == 5120
+assert len(assembly['subject']) == 5120
+assert len(assembly['session']) == 5120
+
+
+# # make sure there are 1280 unique images (shown 1 time for each  of 4 subjects, total of 4 * 1280 = 5120 images shown)
+assert len(np.unique(assembly['image_id'].values)) == 1280
+
+# make sure there are 4 unique subjects:
+assert len(np.unique(assembly['subject'].values)) == 4
+
+# make sure there are 16 unique object categories (ground truths):
+assert len(np.unique(assembly['truth'].values)) == 16
+assert len(np.unique(assembly['category'].values)) == 16
+
+
+# upload to S3
+package_data_assembly('brainio_brainscore', assembly, assembly_identifier=assembly.name,
+                      stimulus_set_identifier="Geirhos2021_eidolonIII",
+                      assembly_class="BehavioralAssembly", bucket_name="brainio-brainscore")

brainscore_vision/data/geirhos2021/data_packaging/eidolonIII/eidolonIII_stimulus_set.py

@@ -0,0 +1,82 @@
+from pathlib import Path
+from brainio.stimuli import StimulusSet
+from brainio.packaging import package_stimulus_set
+
+stimuli = []
+image_paths = {}
+stimuli_directory = '../datasets/eidolonIII/dnn/session-1'
+
+
+'''
+Dataset Meta Info (from https://github.com/rgeirhos/generalisation-humans-DNNs)
+
+Sample image from dataset:
+0001_eid_dnn_1-10-10_oven_10_n04111531_17072.png
+
+This is a concatenation of the following information (separated by '_'):
+
+    1) a four-digit number starting with 0000 for the first image in an experiment; 
+       the last image therefore has the number n-1 if n is the number of images in a certain experiment
+    2) short code for experiment name, e.g. 'eid' for eidolon-experiment
+    3) either e.g. 's01' for 'subject-01', or 'dnn' for DNNs
+    4) condition
+    5) category (ground truth)
+    6) a number (just ignore it)
+    7) image identifier in the form a_b.JPEG (or a_b.png), with a being the 
+       WNID (WordNet ID) of the corresponding synset and b being an integer.
+'''
+
+for filepath in Path(stimuli_directory).glob('*.png'):
+
+    # entire name of image file:
+    image_id = filepath.stem
+    image_id_long = image_id
+    split_name = filepath.stem.split('_')
+
+    # ensure proper metadata length per image in set
+    assert len(split_name) == 8
+
+    # Dataset image data, 1-7 from above:
+    image_number = split_name[0]
+    experiment_code = split_name[1]
+    subject = split_name[2]
+    condition = split_name[3]
+    category_ground_truth = split_name[4]
+    random_number = split_name[5]
+
+    # note the split, for a total of 8 metadata fields:
+    wordnet_a = split_name[6]
+    wordnet_b = split_name[7]
+
+    # image lookup ID, same as data assembly. This is the exact image shown to participant.
+    # This is needed, as the raw images have "dnn" in the subject field, even when a human subject
+    # was used. Otherwise, the image name data in the raw data table and the image names themselves are the same.
+    image_lookup_id = "_".join(split_name[3:])
+
+    image_paths[image_id] = filepath
+    stimuli.append({
+        'image_id': image_lookup_id,
+        'image_id_long': image_id_long,
+        'image_number': image_number,
+        'experiment_code': experiment_code,
+        'condition': condition,
+        'truth': category_ground_truth,
+        'category_ground_truth': category_ground_truth,
+        'random_number': random_number,
+        'wordnet_a': wordnet_a,
+        'wordnet_b': wordnet_b,
+    })
+
+stimuli = StimulusSet(stimuli)
+image_id_to_lookup = dict(zip(stimuli['image_id_long'], stimuli['image_id']))
+stimuli.image_paths = image_paths
+stimuli.image_paths = {image_id_to_lookup[image_id]: path
+                       for image_id, path in stimuli.image_paths.items()}
+stimuli.name = 'Geirhos2021_eidolonIII'  # give the StimulusSet an identifier name
+
+# Ensure 1280 images in dataset
+assert len(stimuli) == 1280
+
+# upload to S3
+package_stimulus_set("brainio_brainscore", stimuli, stimulus_set_identifier=stimuli.name,
+                     bucket_name="brainio-brainscore")

brainscore_vision/data/geirhos2021/data_packaging/false-colour/false-colour_data_assembly.py

@@ -0,0 +1,83 @@
+import numpy as np
+
+from brainio.assemblies import BehavioralAssembly
+from brainio.packaging import package_data_assembly
+import pandas as pd
+
+'''
+Experiment Information:
+https://arxiv.org/pdf/1706.06969.pdf
+
+ - 4 subjects 
+ - 1120 images each
+ - 4480 total images shown 
+ - match to sample task, 16AFC
+ - 16 image categories
+ - for the this benchmark (false-colour) subjects saw the EXACT image indicated with the variable/column name
+   image_lookup_id, and not a variation of it (no distortions, editing, etc). Condition is true or false, 
+   based on whether or not the image was presented in false color (True = presented in false color).
+'''
+
+# initial csv to dataframe processing:
+subject_1 = pd.read_csv('data_assemblies/false-colour_subject-01_session_1.csv')
+subject_2 = pd.read_csv('data_assemblies/false-colour_subject-02_session_1.csv')
+subject_3 = pd.read_csv('data_assemblies/false-colour_subject-03_session_1.csv')
+subject_4 = pd.read_csv('data_assemblies/false-colour_subject-04_session_1.csv')
+
+all_subjects = pd.concat([subject_1, subject_2, subject_3, subject_4])
+
+# parse df for the image lookup id. This relates the data assembly with the stimulus set.
+split_cols = all_subjects['imagename'].str.split("_", expand=True)
+drop_cols = split_cols.drop(split_cols.columns[[0, 1, 2]], axis=1)
+all_subjects['image_lookup_id'] = drop_cols.agg("_".join, axis=1).str.replace(".png", "")
+
+
+# construct the assembly
+assembly = BehavioralAssembly(all_subjects['object_response'],
+                              coords={
+                                  'image_id': ('presentation', all_subjects['image_lookup_id']),
+                                  'image_id_long': ('presentation', all_subjects['imagename']),
+                                  'truth': ('presentation', all_subjects['category']),
+                                  'choice': ('presentation', all_subjects['object_response']),
+                                  'category': ('presentation', all_subjects['category']),
+                                  'response_time': ('presentation', all_subjects['rt']),
+                                  'trial': ('presentation', all_subjects['trial']),
+                                  'subject': ('presentation', all_subjects['subj']),
+                                  'session': ('presentation', all_subjects['Session']),
+                              },
+                              dims=['presentation']
+                              )
+
+# give the assembly an identifier name
+assembly.name = 'brendel.Geirhos2021_false-colour'
+
+# make sure assembly dim is correct length
+assert len(assembly['presentation']) == 4480
+
+# make sure assembly coords are correct length
+assert len(assembly['image_id']) == 4480
+assert len(assembly['image_id_long']) == 4480
+assert len(assembly['truth']) == 4480
+assert len(assembly['category']) == 4480
+assert len(assembly['response_time']) == 4480
+assert len(assembly['trial']) == 4480
+assert len(assembly['subject']) == 4480
+assert len(assembly['session']) == 4480
+
+
+# make sure there are 1120 unique images (shown 1 time for each  of 4 subjects, total of 4 * 1120 = 4480 images shown)
+assert len(np.unique(assembly['image_id'].values)) == 1120
+
+# make sure there are 4 unique subjects:
+assert len(np.unique(assembly['subject'].values)) == 4
+
+
+# make sure there are 16 unique object categories (ground truths):
+assert len(np.unique(assembly['truth'].values)) == 16
+assert len(np.unique(assembly['category'].values)) == 16
+
+
+# upload to S3
+package_data_assembly('brainio_brainscore', assembly, assembly_identifier=assembly.name,
+                      stimulus_set_identifier="Geirhos2021_false-colour",
+                      assembly_class="BehavioralAssembly", bucket_name="brainio-brainscore")