Adapt most code to use ScanpathFixations instead of FixationTrains

Signed-off-by: Matthias Kümmerer <[email protected]>

notebooks/Tutorial.ipynb

@@ -344,7 +344,7 @@
    ],
    "source": [
     "rst = np.random.RandomState(seed=23)\n",
-    "scanpath_indices = rst.randint(len(fixations.train_xs), size=3)\n",
+    "scanpath_indices = rst.randint(len(fixations.scanpaths), size=3)\n",
     "\n",
     "for scanpath_index in scanpath_indices:\n",
     "    print(f\"Scanpath no {scanpath_index}:\")\n",
@@ -402,7 +402,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 15,
+   "execution_count": 12,
    "id": "d7a92d74-4512-42c7-b21f-e62e211df361",
    "metadata": {},
    "outputs": [
@@ -446,7 +446,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 17,
+   "execution_count": 13,
    "id": "f2085fb4-6a30-426a-8a47-0da876d92ef7",
    "metadata": {},
    "outputs": [
@@ -472,7 +472,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 19,
+   "execution_count": 14,
    "id": "658a06b0-f339-41d7-9f74-bba4ee9bebb3",
    "metadata": {},
    "outputs": [
@@ -519,7 +519,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 20,
+   "execution_count": 15,
    "id": "a8068b3c-c1ca-4a64-adae-a7ac769e3a82",
    "metadata": {},
    "outputs": [
@@ -547,7 +547,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 21,
+   "execution_count": 16,
    "id": "106515c4-3f4c-43f6-8d58-501045132c0f",
    "metadata": {},
    "outputs": [],
@@ -593,7 +593,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 22,
+   "execution_count": 17,
    "id": "27fbe076-72d3-4441-a095-69a455e2a210",
    "metadata": {},
    "outputs": [
@@ -634,7 +634,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 23,
+   "execution_count": 18,
    "id": "b1b6dad7-e6a9-4a66-b924-8dbc3312ec94",
    "metadata": {},
    "outputs": [
@@ -667,7 +667,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 24,
+   "execution_count": 19,
    "id": "b66e64ed-ff99-4932-98f8-35cabf60c26c",
    "metadata": {},
    "outputs": [
@@ -712,7 +712,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 25,
+   "execution_count": 20,
    "id": "b20ad889-dbeb-41d6-b79d-1adb0b23150a",
    "metadata": {},
    "outputs": [
@@ -783,7 +783,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 28,
+   "execution_count": 21,
    "id": "9696744b-6cf7-4c43-8fe8-93a6f1a0fa4d",
    "metadata": {},
    "outputs": [],
@@ -842,7 +842,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 29,
+   "execution_count": 22,
    "id": "3f010b8c-d368-4c5c-b241-1ebf88219b94",
    "metadata": {},
    "outputs": [
@@ -889,7 +889,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 30,
+   "execution_count": 23,
    "id": "e74fe712-34fa-4201-8415-8542c0080ffc",
    "metadata": {},
    "outputs": [],
@@ -922,7 +922,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 32,
+   "execution_count": 24,
    "id": "38220bad-2512-465d-8d90-f4bfc7426667",
    "metadata": {},
    "outputs": [
@@ -935,7 +935,7 @@
        "        1.67418705,  3.50254012,  3.13763833,  2.68713872,  1.0400994 ])"
       ]
      },
-     "execution_count": 32,
+     "execution_count": 24,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -963,7 +963,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 33,
+   "execution_count": 25,
    "id": "9e032691-527e-456e-8639-7f360e4582e3",
    "metadata": {},
    "outputs": [
@@ -991,7 +991,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 34,
+   "execution_count": 26,
    "id": "442b8e2e-24a6-44c0-9992-e37336317962",
    "metadata": {},
    "outputs": [
@@ -1037,7 +1037,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 35,
+   "execution_count": 27,
    "id": "3255672f-e3df-4b1d-8cad-cb8e7a39fcd2",
    "metadata": {},
    "outputs": [],
@@ -1047,7 +1047,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 36,
+   "execution_count": 28,
    "id": "f8c137e6-0fa0-4f19-bf92-75e93ed6bbab",
    "metadata": {},
    "outputs": [
@@ -1112,7 +1112,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 37,
+   "execution_count": 29,
    "id": "c17d068c-96c2-410b-8a62-fc229ad45453",
    "metadata": {},
    "outputs": [
@@ -1141,7 +1141,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 38,
+   "execution_count": 30,
    "id": "87c10abe-57a2-4bbc-937f-74488a2636e3",
    "metadata": {},
    "outputs": [

pysaliency/baseline_utils.py

@@ -4,7 +4,7 @@
 import numba
 import numpy as np
 from boltons.iterutils import chunked
-from scipy.ndimage.filters import gaussian_filter
+from scipy.ndimage import gaussian_filter
 from scipy.special import logsumexp
 from sklearn.base import BaseEstimator, DensityMixin
 from sklearn.model_selection import cross_val_score
@@ -137,7 +137,7 @@ def __iter__(self):
         for n in range(len(self.stimuli)):
             for s in range(self.fixations.subject_count):
                 image_inds = self.fixations.n == n
-                subject_inds = self.fixations.subjects == s
+                subject_inds = self.fixations.subject == s
                 train_inds, test_inds = image_inds & ~subject_inds, image_inds & subject_inds
                 if test_inds.sum() == 0 or train_inds.sum() == 0:
                     #print("Skipping")
@@ -152,7 +152,7 @@ def __iter__(self):
                 yield train_inds, test_inds
 
     def __len__(self):
-        return len(set(zip(self.fixations.n, self.fixations.subjects)))
+        return len(set(zip(self.fixations.n, self.fixations.subject)))
 
 
 class ScikitLearnWithinImageCrossValidationGenerator(object):

pysaliency/datasets/__init__.py

@@ -58,7 +58,7 @@ def create_subset(stimuli, fixations, stimuli_indices):
 
         new_image_indices = [new_pos[i] for i in fixations.scanpaths.n[scanpath_inds]]
 
-        new_fixations = fixations.filter_fixation_trains(scanpath_inds)
+        new_fixations = fixations.filter_scanpaths(scanpath_inds)
         new_fixations.scanpaths.n = np.array(new_image_indices)
 
         new_fixation_ns = [new_pos[i] for i in new_fixations.n]

pysaliency/datasets/fixations.py

@@ -6,7 +6,7 @@
 import numpy as np
 from tqdm import tqdm
 
-from ..utils import remove_trailing_nans
+from ..utils import remove_trailing_nans, deprecated_class
 from ..utils.variable_length_array import VariableLengthArray
 from .scanpaths import Scanpaths
 from .utils import _load_attribute_dict_from_hdf5, concatenate_attributes, decode_string, get_merged_attribute_list, hdf5_wrapper
@@ -315,7 +315,10 @@ def to_hdf5(self, target):
         variable_length_arrays = []
 
         for attribute in ['x', 'y', 't', 'x_hist', 'y_hist', 't_hist', 'n', 'lengths'] + self.__attributes__:
-            data = getattr(self, attribute)
+            if attribute == 'lengths':
+                data = self.scanpath_history_length
+            else:
+                data = getattr(self, attribute)
             if isinstance(data, VariableLengthArray):
                 variable_length_arrays.append(attribute)
                 data = data._data
@@ -528,6 +531,7 @@ def read_hdf5(cls, source):
         return cls(scanpaths=scanpaths)
 
 
+@deprecated_class(deprecated_in="0.3.0", removed_in="1.0.0", details="Use `ScanpathFixations` instead")
 class FixationTrains(ScanpathFixations):
     """
     Capsules the fixations of a dataset as fixation trains.
@@ -552,6 +556,8 @@ class FixationTrains(ScanpathFixations):
     """
     def __init__(self, train_xs, train_ys, train_ts, train_ns, train_subjects, scanpath_attributes=None, scanpath_fixation_attributes=None, attributes=None, scanpath_attribute_mapping=None, scanpaths=None):
 
+        # raise ValueError("DON'T USE FIXATIONTRAINS ANYMORE, USE SCANPATHFIXATIONS INSTEAD")
+
         if isinstance(train_xs, Scanpaths):
             scanpaths = train_xs
 
@@ -580,7 +586,7 @@ def __init__(self, train_xs, train_ys, train_ts, train_ns, train_subjects, scanp
             attributes = {}
 
         if attributes:
-            warnings.warn("don't use attributes for FixationTrains, use scanpath_attributes or scanpath_fixation_attributes instead!", stacklevel=2)
+            warnings.warn("Don't use attributes for FixationTrains, use scanpath_attributes or scanpath_fixation_attributes instead! FixationTrains is deprecated, the successor ScanpathFixations doesn't support attributes anymore", stacklevel=2, category=DeprecationWarning)
 
         if attributes:
             self.__attributes__ = list(self.__attributes__)

pysaliency/datasets/scanpaths.py

@@ -33,7 +33,7 @@ def __init__(self,
                  length=None,
                  scanpath_attributes: Optional[Dict[str, np.ndarray]] = None,
                  fixation_attributes: Optional[Dict[str, Union[np.ndarray, VariableLengthArray]]]=None,
-                 attribute_mapping=Dict[str, str],
+                 attribute_mapping: Optional[Dict[str, str]] = None,
                  **kwargs):
 
         self.n = np.asarray(n)
@@ -57,6 +57,7 @@ def __init__(self,
         scanpath_attributes = scanpath_attributes or {}
         fixation_attributes = fixation_attributes or {}
         self.attribute_mapping = attribute_mapping or {}
+        self.attribute_mapping = dict(self.attribute_mapping)
 
 
         for key, value in kwargs.items():

pysaliency/external_datasets/mit.py

@@ -106,7 +106,7 @@ def check_size(f):
                     subject_path = os.path.join('DATA', subject)
                     outfile = '{0}_{1}.mat'.format(stimulus, subject)
                     outfile = os.path.join(out_path, outfile)
-                    cmds.append("fprintf('%d/%d\\r', {}, {});".format(n * len(subjects) + subject_id, total_cmd_count))
+                    cmds.append("fprintf('Processing scanpath %d/%d\\r', {}, {});".format(n * len(subjects) + subject_id, total_cmd_count))
                     cmds.append("extract_fixations('{0}', '{1}', '{2}');".format(stimulus, subject_path, outfile))
 
             print('Running original code to extract fixations. This can take some minutes.')

pysaliency/filter_datasets.py

@@ -4,7 +4,7 @@
 
 from boltons.iterutils import chunked
 
-from .datasets import create_subset, FixationTrains, Fixations, Stimuli
+from .datasets import create_subset, FixationTrains, Fixations, Stimuli, ScanpathFixations
 
 
 def train_split(stimuli, fixations, crossval_folds, fold_no, val_folds=1, test_folds=1, random=True, stratified_attributes=None):
@@ -234,17 +234,17 @@ def filter_stimuli_by_size(stimuli, fixations, size=None, sizes=None):
     return create_subset(stimuli, fixations, indices)
 
 
-def filter_scanpaths_by_attribute(scanpaths:  FixationTrains, attribute_name, attribute_value, invert_match=False):
-    """Filter Scanpaths by values of scanpath attribute (fixation_trains.scanpath_attributes)"""
+def filter_scanpaths_by_attribute(scanpaths: ScanpathFixations, attribute_name, attribute_value, invert_match=False):
+    """Filter Scanpaths by values of scanpath attribute (fixation_trains.scanpaths.scanpath_attributes)"""
 
-    mask = scanpaths.scanpath_attributes[attribute_name] == attribute_value
+    mask = scanpaths.scanpaths.scanpath_attributes[attribute_name] == attribute_value
     if mask.ndim > 1:
         mask = np.all(mask, axis=1)
 
     if invert_match is True:
         mask = ~mask
 
-    return scanpaths.filter_fixation_trains(mask)
+    return scanpaths.filter_scanpaths(mask)
 
 
 def filter_fixations_by_attribute(fixations: Fixations, attribute_name, attribute_value, invert_match=False):
@@ -280,20 +280,20 @@ def filter_stimuli_by_attribute(stimuli: Stimuli, fixations: Fixations, attribut
     return create_subset(stimuli, fixations, indices)
 
 
-def filter_scanpaths_by_length(scanpaths: FixationTrains, intervals: list):
+def filter_scanpaths_by_length(scanpath_fixations: ScanpathFixations, intervals: list):
     """Filter Scanpaths by number of fixations"""
 
     intervals = _check_intervals(intervals, type=int)
-    mask = np.zeros(len(scanpaths.train_lengths), dtype=bool)
+    mask = np.zeros(len(scanpath_fixations.scanpaths), dtype=bool)
     for start, end in intervals:
         temp_mask = np.logical_and(
-            scanpaths.train_lengths >= start, scanpaths.train_lengths < end)
+            scanpath_fixations.scanpaths.length >= start, scanpath_fixations.scanpaths.length < end)
         mask = np.logical_or(mask, temp_mask)
     indices = list(np.nonzero(mask)[0])
 
-    scanpaths = scanpaths.filter_fixation_trains(indices)
+    scanpath_fixations = scanpath_fixations.filter_scanpaths(indices)
 
-    return scanpaths
+    return scanpath_fixations
 
 
 def remove_stimuli_without_fixations(stimuli: Stimuli, fixations: Fixations):

pysaliency/models.py

@@ -16,7 +16,7 @@
                                   DisjointUnionMixin,
                                   GaussianSaliencyMapModel,
                                   )
-from .datasets import FixationTrains, check_prediction_shape, get_image_hash, as_stimulus
+from .datasets import Scanpaths, ScanpathFixations, check_prediction_shape, get_image_hash, as_stimulus
 from .metrics import probabilistic_image_based_kl_divergence, convert_saliency_map_to_density
 from .sampling_models import SamplingModelMixin
 from .utils import Cache, average_values, deprecated_class, remove_trailing_nans, iterator_chunks
@@ -212,7 +212,7 @@ def _expand_sample_arguments(self, stimuli, train_counts, lengths=None, stimulus
 
         return stimuli, train_counts, lengths, stimulus_indices
 
-    def sample(self, stimuli, train_counts, lengths=1, stimulus_indices=None, rst=None, verbose=False):
+    def sample(self, stimuli, train_counts, lengths=1, stimulus_indices=None, rst=None, verbose=False) -> ScanpathFixations:
         """
         Sample fixations for given stimuli
 
@@ -261,7 +261,7 @@ def sample(self, stimuli, train_counts, lengths=1, stimulus_indices=None, rst=No
                 ns.append(stimulus_index)
                 subjects.append(0)
                 pbar.update(1)
-        return FixationTrains.from_fixation_trains(xs, ys, ts, ns, subjects)
+        return ScanpathFixations(Scanpaths(xs=xs, ys=ys, ts=ts, n=ns, subject=subjects))
 
     def _sample_fixation_train(self, stimulus, length, rst=None):
         """Sample one fixation train of given length from stimulus"""

pysaliency/saliency_map_models.py

@@ -956,12 +956,12 @@ def __init__(self, subject_models, **kwargs):
 
     def _split_fixations(self, stimuli, fixations):
         for s in self.subject_models:
-            yield fixations.subjects == s, self.subject_models[s]
+            yield fixations.subject == s, self.subject_models[s]
 
     def conditional_saliency_map(self, stimulus, x_hist, y_hist, t_hist, attributes=None, out=None, **kwargs):
-        if 'subjects' not in attributes:
+        if 'subject' not in attributes:
             raise ValueError("SubjectDependentSaliencyModel can't compute conditional saliency maps without subject indication!")
-        return self.subject_models[attributes['subjects']].conditional_saliency_map(
+        return self.subject_models[attributes['subject']].conditional_saliency_map(
             stimulus, x_hist, y_hist, t_hist, attributes=attributes, **kwargs)