interpret: update normlime

paddlex/interpret/core/_session_preparation.py

@@ -20,6 +20,7 @@ import numpy as np
 from paddle.fluid.param_attr import ParamAttr
 from paddlex.interpret.as_data_reader.readers import preprocess_image
 
+
 def gen_user_home():
     if "HOME" in os.environ:
         home_path = os.environ["HOME"]
@@ -34,10 +35,20 @@ def paddle_get_fc_weights(var_name="fc_0.w_0"):
 
 
 def paddle_resize(extracted_features, outsize):
-    resized_features = fluid.layers.resize_bilinear(extracted_features, outsize)
+    resized_features = fluid.layers.resize_bilinear(extracted_features,
+                                                    outsize)
     return resized_features
 
 
+def get_precomputed_normlime_weights():
+    root_path = gen_user_home()
+    root_path = osp.join(root_path, '.paddlex')
+    h_pre_models = osp.join(root_path, "pre_models")
+    normlime_weights_file = osp.join(
+        h_pre_models, "normlime_weights_imagenet_resnet50vc.npy")
+    return np.load(normlime_weights_file, allow_pickle=True).item()
+
+
 def compute_features_for_kmeans(data_content):
     root_path = gen_user_home()
     root_path = osp.join(root_path, '.paddlex')
@@ -47,6 +58,7 @@ def compute_features_for_kmeans(data_content):
             os.makedirs(root_path)
         url = "https://bj.bcebos.com/paddlex/interpret/pre_models.tar.gz"
         pdx.utils.download_and_decompress(url, path=root_path)
+
     def conv_bn_layer(input,
                       num_filters,
                       filter_size,
@@ -55,7 +67,7 @@ def compute_features_for_kmeans(data_content):
                       act=None,
                       name=None,
                       is_test=True,
-                      global_name=''):
+                      global_name='for_kmeans_'):
         conv = fluid.layers.conv2d(
             input=input,
             num_filters=num_filters,
@@ -79,14 +91,14 @@ def compute_features_for_kmeans(data_content):
             bias_attr=ParamAttr(global_name + bn_name + '_offset'),
             moving_mean_name=global_name + bn_name + '_mean',
             moving_variance_name=global_name + bn_name + '_variance',
-            use_global_stats=is_test
-        )
+            use_global_stats=is_test)
 
     startup_prog = fluid.default_startup_program().clone(for_test=True)
     prog = fluid.Program()
     with fluid.program_guard(prog, startup_prog):
         with fluid.unique_name.guard():
-            image_op = fluid.data(name='image', shape=[None, 3, 224, 224], dtype='float32')
+            image_op = fluid.data(
+                name='image', shape=[None, 3, 224, 224], dtype='float32')
 
             conv = conv_bn_layer(
                 input=image_op,
@@ -110,7 +122,8 @@ def compute_features_for_kmeans(data_content):
                 act='relu',
                 name='conv1_3')
             extracted_features = conv
-            resized_features = fluid.layers.resize_bilinear(extracted_features, image_op.shape[2:])
+            resized_features = fluid.layers.resize_bilinear(extracted_features,
+                                                            image_op.shape[2:])
 
     gpu_id = int(os.environ.get('FLAGS_selected_gpus', 0))
     place = fluid.CUDAPlace(gpu_id)
@@ -119,7 +132,10 @@ def compute_features_for_kmeans(data_content):
     exe.run(startup_prog)
     fluid.io.load_persistables(exe, h_pre_models, prog)
 
-    images = preprocess_image(data_content)  # transpose to [N, 3, H, W], scaled to [0.0, 1.0]
-    result = exe.run(prog, fetch_list=[resized_features], feed={'image': images})
+    images = preprocess_image(
+        data_content)  # transpose to [N, 3, H, W], scaled to [0.0, 1.0]
+    result = exe.run(prog,
+                     fetch_list=[resized_features],
+                     feed={'image': images})
 
     return result[0][0]

paddlex/interpret/core/interpretation.py

@@ -20,12 +20,10 @@ class Interpretation(object):
     """
     Base class for all interpretation algorithms.
     """
-    def __init__(self, interpretation_algorithm_name, predict_fn, label_names, **kwargs):
-        supported_algorithms = {
-            'cam': CAM,
-            'lime': LIME,
-            'normlime': NormLIME
-        }
+
+    def __init__(self, interpretation_algorithm_name, predict_fn, label_names,
+                 **kwargs):
+        supported_algorithms = {'cam': CAM, 'lime': LIME, 'normlime': NormLIME}
 
         self.algorithm_name = interpretation_algorithm_name.lower()
         assert self.algorithm_name in supported_algorithms.keys()
@@ -33,10 +31,13 @@ class Interpretation(object):
 
         # initialization for the interpretation algorithm.
         self.algorithm = supported_algorithms[self.algorithm_name](
-            self.predict_fn, label_names, **kwargs
-        )
+            self.predict_fn, label_names, **kwargs)
 
-    def interpret(self, data_, visualization=True, save_to_disk=True, save_dir='./tmp'):
+    def interpret(self,
+                  data_,
+                  visualization=True,
+                  save_to_disk=True,
+                  save_dir='./tmp'):
         """
 
         Args:
@@ -48,4 +49,5 @@ class Interpretation(object):
         Returns:
 
         """
-        return self.algorithm.interpret(data_, visualization, save_to_disk, save_dir)
+        return self.algorithm.interpret(data_, visualization, save_to_disk,
+                                        save_dir)

paddlex/interpret/core/lime_base.py

@@ -27,7 +27,6 @@ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 The code in this file (lime_base.py) is modified from https://github.com/marcotcr/lime.
 """
 
-
 import numpy as np
 import scipy as sp
 
@@ -39,10 +38,8 @@ import paddlex.utils.logging as logging
 
 class LimeBase(object):
     """Class for learning a locally linear sparse model from perturbed data"""
-    def __init__(self,
-                 kernel_fn,
-                 verbose=False,
-                 random_state=None):
+
+    def __init__(self, kernel_fn, verbose=False, random_state=None):
         """Init function
 
         Args:
@@ -72,15 +69,14 @@ class LimeBase(object):
         """
         from sklearn.linear_model import lars_path
         x_vector = weighted_data
-        alphas, _, coefs = lars_path(x_vector,
-                                     weighted_labels,
-                                     method='lasso',
-                                     verbose=False)
+        alphas, _, coefs = lars_path(
+            x_vector, weighted_labels, method='lasso', verbose=False)
         return alphas, coefs
 
     def forward_selection(self, data, labels, weights, num_features):
         """Iteratively adds features to the model"""
-        clf = Ridge(alpha=0, fit_intercept=True, random_state=self.random_state)
+        clf = Ridge(
+            alpha=0, fit_intercept=True, random_state=self.random_state)
         used_features = []
         for _ in range(min(num_features, data.shape[1])):
             max_ = -100000000
@@ -88,11 +84,13 @@ class LimeBase(object):
             for feature in range(data.shape[1]):
                 if feature in used_features:
                     continue
-                clf.fit(data[:, used_features + [feature]], labels,
+                clf.fit(data[:, used_features + [feature]],
+                        labels,
                         sample_weight=weights)
-                score = clf.score(data[:, used_features + [feature]],
-                                  labels,
-                                  sample_weight=weights)
+                score = clf.score(
+                    data[:, used_features + [feature]],
+                    labels,
+                    sample_weight=weights)
                 if score > max_:
                     best = feature
                     max_ = score
@@ -108,8 +106,8 @@ class LimeBase(object):
         elif method == 'forward_selection':
             return self.forward_selection(data, labels, weights, num_features)
         elif method == 'highest_weights':
-            clf = Ridge(alpha=0.01, fit_intercept=True,
-                        random_state=self.random_state)
+            clf = Ridge(
+                alpha=0.01, fit_intercept=True, random_state=self.random_state)
             clf.fit(data, labels, sample_weight=weights)
 
             coef = clf.coef_
@@ -125,7 +123,8 @@ class LimeBase(object):
                     nnz_indexes = argsort_data[::-1]
                     indices = weighted_data.indices[nnz_indexes]
                     num_to_pad = num_features - sdata
-                    indices = np.concatenate((indices, np.zeros(num_to_pad, dtype=indices.dtype)))
+                    indices = np.concatenate((indices, np.zeros(
+                        num_to_pad, dtype=indices.dtype)))
                     indices_set = set(indices)
                     pad_counter = 0
                     for i in range(data.shape[1]):
@@ -135,7 +134,8 @@ class LimeBase(object):
                             if pad_counter >= num_to_pad:
                                 break
                 else:
-                    nnz_indexes = argsort_data[sdata - num_features:sdata][::-1]
+                    nnz_indexes = argsort_data[sdata - num_features:sdata][::
+                                                                           -1]
                     indices = weighted_data.indices[nnz_indexes]
                 return indices
             else:
@@ -146,13 +146,13 @@ class LimeBase(object):
                     reverse=True)
                 return np.array([x[0] for x in feature_weights[:num_features]])
         elif method == 'lasso_path':
-            weighted_data = ((data - np.average(data, axis=0, weights=weights))
-                             * np.sqrt(weights[:, np.newaxis]))
-            weighted_labels = ((labels - np.average(labels, weights=weights))
-                               * np.sqrt(weights))
+            weighted_data = ((data - np.average(
+                data, axis=0, weights=weights)) *
+                             np.sqrt(weights[:, np.newaxis]))
+            weighted_labels = ((labels - np.average(
+                labels, weights=weights)) * np.sqrt(weights))
             nonzero = range(weighted_data.shape[1])
-            _, coefs = self.generate_lars_path(weighted_data,
-                                               weighted_labels)
+            _, coefs = self.generate_lars_path(weighted_data, weighted_labels)
             for i in range(len(coefs.T) - 1, 0, -1):
                 nonzero = coefs.T[i].nonzero()[0]
                 if len(nonzero) <= num_features:
@@ -164,8 +164,8 @@ class LimeBase(object):
                 n_method = 'forward_selection'
             else:
                 n_method = 'highest_weights'
-            return self.feature_selection(data, labels, weights,
-                                          num_features, n_method)
+            return self.feature_selection(data, labels, weights, num_features,
+                                          n_method)
 
     def interpret_instance_with_data(self,
                                      neighborhood_data,
@@ -214,30 +214,31 @@ class LimeBase(object):
         weights = self.kernel_fn(distances)
         labels_column = neighborhood_labels[:, label]
         used_features = self.feature_selection(neighborhood_data,
-                                               labels_column,
-                                               weights,
-                                               num_features,
-                                               feature_selection)
+                                               labels_column, weights,
+                                               num_features, feature_selection)
         if model_regressor is None:
-            model_regressor = Ridge(alpha=1, fit_intercept=True,
-                                    random_state=self.random_state)
+            model_regressor = Ridge(
+                alpha=1, fit_intercept=True, random_state=self.random_state)
         easy_model = model_regressor
         easy_model.fit(neighborhood_data[:, used_features],
-                       labels_column, sample_weight=weights)
+                       labels_column,
+                       sample_weight=weights)
         prediction_score = easy_model.score(
             neighborhood_data[:, used_features],
-            labels_column, sample_weight=weights)
+            labels_column,
+            sample_weight=weights)
 
-        local_pred = easy_model.predict(neighborhood_data[0, used_features].reshape(1, -1))
+        local_pred = easy_model.predict(neighborhood_data[0, used_features]
+                                        .reshape(1, -1))
 
         if self.verbose:
             logging.info('Intercept' + str(easy_model.intercept_))
             logging.info('Prediction_local' + str(local_pred))
             logging.info('Right:' + str(neighborhood_labels[0, label]))
-        return (easy_model.intercept_,
-                sorted(zip(used_features, easy_model.coef_),
-                       key=lambda x: np.abs(x[1]), reverse=True),
-                prediction_score, local_pred)
+        return (easy_model.intercept_, sorted(
+            zip(used_features, easy_model.coef_),
+            key=lambda x: np.abs(x[1]),
+            reverse=True), prediction_score, local_pred)
 
 
 class ImageInterpretation(object):
@@ -254,8 +255,13 @@ class ImageInterpretation(object):
         self.local_weights = {}
         self.local_pred = None
 
-    def get_image_and_mask(self, label, positive_only=True, negative_only=False, hide_rest=False,
-                           num_features=5, min_weight=0.):
+    def get_image_and_mask(self,
+                           label,
+                           positive_only=True,
+                           negative_only=False,
+                           hide_rest=False,
+                           num_features=5,
+                           min_weight=0.):
         """Init function.
 
         Args:
@@ -279,7 +285,9 @@ class ImageInterpretation(object):
         if label not in self.local_weights:
             raise KeyError('Label not in interpretation')
         if positive_only & negative_only:
-            raise ValueError("Positive_only and negative_only cannot be true at the same time.")
+            raise ValueError(
+                "Positive_only and negative_only cannot be true at the same time."
+            )
         segments = self.segments
         image = self.image
         local_weights_label = self.local_weights[label]
@@ -289,14 +297,20 @@ class ImageInterpretation(object):
         else:
             temp = self.image.copy()
         if positive_only:
-            fs = [x[0] for x in local_weights_label
-                  if x[1] > 0 and x[1] > min_weight][:num_features]
+            fs = [
+                x[0] for x in local_weights_label
+                if x[1] > 0 and x[1] > min_weight
+            ][:num_features]
         if negative_only:
-            fs = [x[0] for x in local_weights_label
-                  if x[1] < 0 and abs(x[1]) > min_weight][:num_features]
+            fs = [
+                x[0] for x in local_weights_label
+                if x[1] < 0 and abs(x[1]) > min_weight
+            ][:num_features]
         if positive_only or negative_only:
+            c = 1 if positive_only else 0
             for f in fs:
-                temp[segments == f] = image[segments == f].copy()
+                temp[segments == f] = [0, 255, 0]
+                # temp[segments == f, c] = np.max(image)
                 mask[segments == f] = 1
             return temp, mask
         else:
@@ -330,8 +344,11 @@ class ImageInterpretation(object):
         temp = np.zeros_like(image)
 
         weight_max = abs(local_weights_label[0][1])
-        local_weights_label = [(f, w/weight_max) for f, w in local_weights_label]
-        local_weights_label = sorted(local_weights_label, key=lambda x: x[1], reverse=True)  # negatives are at last.
+        local_weights_label = [(f, w / weight_max)
+                               for f, w in local_weights_label]
+        local_weights_label = sorted(
+            local_weights_label, key=lambda x: x[1],
+            reverse=True)  # negatives are at last.
 
         cmaps = cm.get_cmap('Spectral')
         colors = cmaps(np.linspace(0, 1, len(local_weights_label)))
@@ -354,8 +371,12 @@ class LimeImageInterpreter(object):
     feature that is 1 when the value is the same as the instance being
     interpreted."""
 
-    def __init__(self, kernel_width=.25, kernel=None, verbose=False,
-                 feature_selection='auto', random_state=None):
+    def __init__(self,
+                 kernel_width=.25,
+                 kernel=None,
+                 verbose=False,
+                 feature_selection='auto',
+                 random_state=None):
         """Init function.
 
         Args:
@@ -377,22 +398,27 @@ class LimeImageInterpreter(object):
         kernel_width = float(kernel_width)
 
         if kernel is None:
+
             def kernel(d, kernel_width):
-                return np.sqrt(np.exp(-(d ** 2) / kernel_width ** 2))
+                return np.sqrt(np.exp(-(d**2) / kernel_width**2))
 
         kernel_fn = partial(kernel, kernel_width=kernel_width)
 
         self.random_state = check_random_state(random_state)
         self.feature_selection = feature_selection
-        self.base = LimeBase(kernel_fn, verbose, random_state=self.random_state)
+        self.base = LimeBase(
+            kernel_fn, verbose, random_state=self.random_state)
 
-    def interpret_instance(self, image, classifier_fn, labels=(1,),
+    def interpret_instance(self,
+                           image,
+                           classifier_fn,
+                           labels=(1, ),
                            hide_color=None,
-                           num_features=100000, num_samples=1000,
+                           num_features=100000,
+                           num_samples=1000,
                            batch_size=10,
                            distance_metric='cosine',
-                           model_regressor=None
-                           ):
+                           model_regressor=None):
         """Generates interpretations for a prediction.
 
         First, we generate neighborhood data by randomly perturbing features
@@ -435,6 +461,7 @@ class LimeImageInterpreter(object):
         self.segments = segments
 
         fudged_image = image.copy()
+        # global_mean = np.mean(image, (0, 1))
         if hide_color is None:
             # if no hide_color, use the mean
             for x in np.unique(segments):
@@ -461,24 +488,30 @@ class LimeImageInterpreter(object):
 
         top = labels
 
-        data, labels = self.data_labels(image, fudged_image, segments,
-                                        classifier_fn, num_samples,
-                                        batch_size=batch_size)
+        data, labels = self.data_labels(
+            image,
+            fudged_image,
+            segments,
+            classifier_fn,
+            num_samples,
+            batch_size=batch_size)
 
         distances = sklearn.metrics.pairwise_distances(
-            data,
-            data[0].reshape(1, -1),
-            metric=distance_metric
-        ).ravel()
+            data, data[0].reshape(1, -1), metric=distance_metric).ravel()
 
         interpretation_image = ImageInterpretation(image, segments)
         for label in top:
             (interpretation_image.intercept[label],
              interpretation_image.local_weights[label],
-             interpretation_image.score, interpretation_image.local_pred) = self.base.interpret_instance_with_data(
-                data, labels, distances, label, num_features,
-                model_regressor=model_regressor,
-                feature_selection=self.feature_selection)
+             interpretation_image.score, interpretation_image.local_pred
+             ) = self.base.interpret_instance_with_data(
+                 data,
+                 labels,
+                 distances,
+                 label,
+                 num_features,
+                 model_regressor=model_regressor,
+                 feature_selection=self.feature_selection)
         return interpretation_image
 
     def data_labels(self,

paddlex/interpret/core/normlime_base.py

@@ -16,6 +16,7 @@ import os
 import os.path as osp
 import numpy as np
 import glob
+import tqdm
 
 from paddlex.interpret.as_data_reader.readers import read_image
 import paddlex.utils.logging as logging
@@ -38,18 +39,24 @@ def combine_normlime_and_lime(lime_weights, g_weights):
 
     for y in pred_labels:
         normlized_lime_weights_y = lime_weights[y]
-        lime_weights_dict = {tuple_w[0]: tuple_w[1] for tuple_w in normlized_lime_weights_y}
+        lime_weights_dict = {
+            tuple_w[0]: tuple_w[1]
+            for tuple_w in normlized_lime_weights_y
+        }
 
         normlized_g_weight_y = g_weights[y]
-        normlime_weights_dict = {tuple_w[0]: tuple_w[1] for tuple_w in normlized_g_weight_y}
+        normlime_weights_dict = {
+            tuple_w[0]: tuple_w[1]
+            for tuple_w in normlized_g_weight_y
+        }
 
         combined_weights[y] = [
             (seg_k, lime_weights_dict[seg_k] * normlime_weights_dict[seg_k])
             for seg_k in lime_weights_dict.keys()
         ]
 
-        combined_weights[y] = sorted(combined_weights[y],
-                                     key=lambda x: np.abs(x[1]), reverse=True)
+        combined_weights[y] = sorted(
+            combined_weights[y], key=lambda x: np.abs(x[1]), reverse=True)
 
     return combined_weights
 
@@ -67,7 +74,8 @@ def centroid_using_superpixels(features, segments):
     regions = regionprops(segments + 1)
     one_list = np.zeros((len(np.unique(segments)), features.shape[2]))
     for i, r in enumerate(regions):
-        one_list[i] = features[int(r.centroid[0] + 0.5), int(r.centroid[1] + 0.5), :]
+        one_list[i] = features[int(r.centroid[0] + 0.5), int(r.centroid[1] +
+                                                             0.5), :]
     return one_list
 
 
@@ -80,30 +88,39 @@ def get_feature_for_kmeans(feature_map, segments):
     return x
 
 
-def precompute_normlime_weights(list_data_, predict_fn, num_samples=3000, batch_size=50, save_dir='./tmp'):
+def precompute_normlime_weights(list_data_,
+                                predict_fn,
+                                num_samples=3000,
+                                batch_size=50,
+                                save_dir='./tmp'):
     # save lime weights and kmeans cluster labels
-    precompute_lime_weights(list_data_, predict_fn, num_samples, batch_size, save_dir)
+    precompute_lime_weights(list_data_, predict_fn, num_samples, batch_size,
+                            save_dir)
 
     # load precomputed results, compute normlime weights and save.
-    fname_list = glob.glob(os.path.join(save_dir, 'lime_weights_s{}*.npy'.format(num_samples)))
+    fname_list = glob.glob(
+        os.path.join(save_dir, 'lime_weights_s{}*.npy'.format(num_samples)))
     return compute_normlime_weights(fname_list, save_dir, num_samples)
 
 
-def save_one_lime_predict_and_kmean_labels(lime_all_weights, image_pred_labels, cluster_labels, save_path):
+def save_one_lime_predict_and_kmean_labels(lime_all_weights, image_pred_labels,
+                                           cluster_labels, save_path):
 
     lime_weights = {}
     for label in image_pred_labels:
         lime_weights[label] = lime_all_weights[label]
 
     for_normlime_weights = {
-        'lime_weights': lime_weights,  # a dict: class_label: (seg_label, weight)
+        'lime_weights':
+        lime_weights,  # a dict: class_label: (seg_label, weight)
         'cluster': cluster_labels  # a list with segments as indices.
     }
 
     np.save(save_path, for_normlime_weights)
 
 
-def precompute_lime_weights(list_data_, predict_fn, num_samples, batch_size, save_dir):
+def precompute_lime_weights(list_data_, predict_fn, num_samples, batch_size,
+                            save_dir):
     root_path = gen_user_home()
     root_path = osp.join(root_path, '.paddlex')
     h_pre_models = osp.join(root_path, "pre_models")
@@ -117,17 +134,24 @@ def precompute_lime_weights(list_data_, predict_fn, num_samples, batch_size, sav
 
     for data_index, each_data_ in enumerate(list_data_):
         if isinstance(each_data_, str):
-            save_path = "lime_weights_s{}_{}.npy".format(num_samples, each_data_.split('/')[-1].split('.')[0])
+            save_path = "lime_weights_s{}_{}.npy".format(
+                num_samples, each_data_.split('/')[-1].split('.')[0])
             save_path = os.path.join(save_dir, save_path)
         else:
-            save_path = "lime_weights_s{}_{}.npy".format(num_samples, data_index)
+            save_path = "lime_weights_s{}_{}.npy".format(num_samples,
+                                                         data_index)
             save_path = os.path.join(save_dir, save_path)
 
         if os.path.exists(save_path):
-            logging.info(save_path + ' exists, not computing this one.', use_color=True)
+            logging.info(
+                save_path + ' exists, not computing this one.', use_color=True)
             continue
-        img_file_name = each_data_ if isinstance(each_data_, str) else data_index
-        logging.info('processing '+ img_file_name + ' [{}/{}]'.format(data_index, len(list_data_)), use_color=True)
+        img_file_name = each_data_ if isinstance(each_data_,
+                                                 str) else data_index
+        logging.info(
+            'processing ' + img_file_name + ' [{}/{}]'.format(data_index,
+                                                              len(list_data_)),
+            use_color=True)
 
         image_show = read_image(each_data_)
         result = predict_fn(image_show)
@@ -156,32 +180,38 @@ def precompute_lime_weights(list_data_, predict_fn, num_samples, batch_size, sav
         pred_label = pred_label[:top_k]
 
         algo = lime_base.LimeImageInterpreter()
-        interpreter = algo.interpret_instance(image_show[0], predict_fn, pred_label, 0,
-                                          num_samples=num_samples, batch_size=batch_size)
-
-        X = get_feature_for_kmeans(compute_features_for_kmeans(image_show).transpose((1, 2, 0)), interpreter.segments)
+        interpreter = algo.interpret_instance(
+            image_show[0],
+            predict_fn,
+            pred_label,
+            0,
+            num_samples=num_samples,
+            batch_size=batch_size)
+
+        X = get_feature_for_kmeans(
+            compute_features_for_kmeans(image_show).transpose((1, 2, 0)),
+            interpreter.segments)
         try:
             cluster_labels = kmeans_model.predict(X)
         except AttributeError:
             from sklearn.metrics import pairwise_distances_argmin_min
-            cluster_labels, _ = pairwise_distances_argmin_min(X, kmeans_model.cluster_centers_)
+            cluster_labels, _ = pairwise_distances_argmin_min(
+                X, kmeans_model.cluster_centers_)
         save_one_lime_predict_and_kmean_labels(
-            interpreter.local_weights, pred_label,
-            cluster_labels,
-            save_path
-        )
+            interpreter.local_weights, pred_label, cluster_labels, save_path)
 
 
 def compute_normlime_weights(a_list_lime_fnames, save_dir, lime_num_samples):
     normlime_weights_all_labels = {}
-    
+
     for f in a_list_lime_fnames:
         try:
             lime_weights_and_cluster = np.load(f, allow_pickle=True).item()
             lime_weights = lime_weights_and_cluster['lime_weights']
             cluster = lime_weights_and_cluster['cluster']
         except:
-            logging.info('When loading precomputed LIME result, skipping' + str(f))
+            logging.info('When loading precomputed LIME result, skipping' +
+                         str(f))
             continue
         logging.info('Loading precomputed LIME result,' + str(f))
         pred_labels = lime_weights.keys()
@@ -203,10 +233,12 @@ def compute_normlime_weights(a_list_lime_fnames, save_dir, lime_num_samples):
     for y in normlime_weights_all_labels:
         normlime_weights = normlime_weights_all_labels.get(y, {})
         for k in normlime_weights:
-            normlime_weights[k] = sum(normlime_weights[k]) / len(normlime_weights[k])
+            normlime_weights[k] = sum(normlime_weights[k]) / len(
+                normlime_weights[k])
 
     # check normlime
-    if len(normlime_weights_all_labels.keys()) < max(normlime_weights_all_labels.keys()) + 1:
+    if len(normlime_weights_all_labels.keys()) < max(
+            normlime_weights_all_labels.keys()) + 1:
         logging.info(
             "\n" + \
             "Warning: !!! \n" + \
@@ -218,17 +250,102 @@ def compute_normlime_weights(a_list_lime_fnames, save_dir, lime_num_samples):
         )
 
     n = 0
-    f_out = 'normlime_weights_s{}_samples_{}-{}.npy'.format(lime_num_samples, len(a_list_lime_fnames), n)
-    while os.path.exists(
-            os.path.join(save_dir, f_out)
-    ):
+    f_out = 'normlime_weights_s{}_samples_{}-{}.npy'.format(
+        lime_num_samples, len(a_list_lime_fnames), n)
+    while os.path.exists(os.path.join(save_dir, f_out)):
         n += 1
-        f_out = 'normlime_weights_s{}_samples_{}-{}.npy'.format(lime_num_samples, len(a_list_lime_fnames), n)
+        f_out = 'normlime_weights_s{}_samples_{}-{}.npy'.format(
+            lime_num_samples, len(a_list_lime_fnames), n)
         continue
 
-    np.save(
-        os.path.join(save_dir, f_out),
-        normlime_weights_all_labels
-    )
+    np.save(os.path.join(save_dir, f_out), normlime_weights_all_labels)
     return os.path.join(save_dir, f_out)
 
+
+def precompute_global_classifier(dataset,
+                                 predict_fn,
+                                 save_path,
+                                 batch_size=50,
+                                 max_num_samples=1000):
+    from sklearn.linear_model import LogisticRegression
+
+    root_path = gen_user_home()
+    root_path = osp.join(root_path, '.paddlex')
+    h_pre_models = osp.join(root_path, "pre_models")
+    if not osp.exists(h_pre_models):
+        if not osp.exists(root_path):
+            os.makedirs(root_path)
+        url = "https://bj.bcebos.com/paddlex/interpret/pre_models.tar.gz"
+        pdx.utils.download_and_decompress(url, path=root_path)
+    h_pre_models_kmeans = osp.join(h_pre_models, "kmeans_model.pkl")
+    kmeans_model = load_kmeans_model(h_pre_models_kmeans)
+
+    image_list = []
+    for item in dataset.file_list:
+        image_list.append(item[0])
+
+    x_data = []
+    y_labels = []
+
+    for each_data_ in tqdm.tqdm(image_list):
+        x_data_i = np.zeros((len(kmeans_model.cluster_centers_)))
+        image_show = read_image(each_data_)
+        result = predict_fn(image_show)
+        result = result[0]  # only one image here.
+        c = compute_features_for_kmeans(image_show).transpose((1, 2, 0))
+
+        segments = np.zeros((image_show.shape[1], image_show.shape[2]),
+                            np.int32)
+        num_blocks = 10
+        height_per_i = segments.shape[0] // num_blocks + 1
+        width_per_i = segments.shape[1] // num_blocks + 1
+
+        for i in range(segments.shape[0]):
+            for j in range(segments.shape[1]):
+                segments[i,
+                         j] = i // height_per_i * num_blocks + j // width_per_i
+
+        # segments = quickshift(image_show[0], sigma=1)
+        X = get_feature_for_kmeans(c, segments)
+
+        try:
+            cluster_labels = kmeans_model.predict(X)
+        except AttributeError:
+            from sklearn.metrics import pairwise_distances_argmin_min
+            cluster_labels, _ = pairwise_distances_argmin_min(
+                X, kmeans_model.cluster_centers_)
+
+        for c in cluster_labels:
+            x_data_i[c] = 1
+
+        # x_data_i /= len(cluster_labels)
+
+        pred_y_i = np.argmax(result)
+        y_labels.append(pred_y_i)
+        x_data.append(x_data_i)
+
+    clf = LogisticRegression(multi_class='multinomial', max_iter=1000)
+    clf.fit(x_data, y_labels)
+
+    num_classes = len(np.unique(y_labels))
+    normlime_weights_all_labels = {}
+
+    for class_index in range(num_classes):
+        w = clf.coef_[class_index]
+
+        # softmax
+        w = w - np.max(w)
+        exp_w = np.exp(w * 10)
+        w = exp_w / np.sum(exp_w)
+
+        normlime_weights_all_labels[class_index] = {
+            i: wi
+            for i, wi in enumerate(w)
+        }
+
+    logging.info("Saving the computed normlime_weights in {}".format(
+        save_path))
+
+    np.save(save_path, normlime_weights_all_labels)
+
+    return save_path

paddlex/interpret/interpretation_predict.py

@@ -13,17 +13,26 @@
 # limitations under the License.
 
 import numpy as np
+import cv2
+import copy
+
 
 def interpretation_predict(model, images):
-    model.arrange_transforms(
-            transforms=model.test_transforms, mode='test')
+    images = images.astype('float32')
+    model.arrange_transforms(transforms=model.test_transforms, mode='test')
+    tmp_transforms = copy.deepcopy(model.test_transforms.transforms)
+    model.test_transforms.transforms = model.test_transforms.transforms[-2:]
+
     new_imgs = []
     for i in range(images.shape[0]):
-        img = images[i]
-        new_imgs.append(model.test_transforms(img)[0])
+        images[i] = cv2.cvtColor(images[i], cv2.COLOR_RGB2BGR)
+        new_imgs.append(model.test_transforms(images[i])[0])
+
     new_imgs = np.array(new_imgs)
-    result = model.exe.run(
-        model.test_prog,
-        feed={'image': new_imgs},
-        fetch_list=list(model.interpretation_feats.values()))
-    return result
\ No newline at end of file
+    out = model.exe.run(model.test_prog,
+                        feed={'image': new_imgs},
+                        fetch_list=list(model.interpretation_feats.values()))
+
+    model.test_transforms.transforms = tmp_transforms
+
+    return out

paddlex/interpret/visualize.py

@@ -20,79 +20,79 @@ import numpy as np
 import paddlex as pdx
 from .interpretation_predict import interpretation_predict
 from .core.interpretation import Interpretation
-from .core.normlime_base import precompute_normlime_weights
+from .core.normlime_base import precompute_global_classifier
 from .core._session_preparation import gen_user_home
-   
-def lime(img_file, 
-         model, 
-         num_samples=3000, 
-         batch_size=50,
-         save_dir='./'):
-    """使用LIME算法将模型预测结果的可解释性可视化。 
-    
+
+
+def lime(img_file, model, num_samples=3000, batch_size=50, save_dir='./'):
+    """使用LIME算法将模型预测结果的可解释性可视化。
+
     LIME表示与模型无关的局部可解释性，可以解释任何模型。LIME的思想是以输入样本为中心，
     在其附近的空间中进行随机采样，每个采样通过原模型得到新的输出，这样得到一系列的输入
     和对应的输出，LIME用一个简单的、可解释的模型（比如线性回归模型）来拟合这个映射关系，
-    得到每个输入维度的权重，以此来解释模型。  
-    
+    得到每个输入维度的权重，以此来解释模型。
+
     注意：LIME可解释性结果可视化目前只支持分类模型。
-         
+
     Args:
         img_file (str): 预测图像路径。
         model (paddlex.cv.models): paddlex中的模型。
         num_samples (int): LIME用于学习线性模型的采样数，默认为3000。
         batch_size (int): 预测数据batch大小，默认为50。
-        save_dir (str): 可解释性可视化结果（保存为png格式文件）和中间文件存储路径。        
+        save_dir (str): 可解释性可视化结果（保存为png格式文件）和中间文件存储路径。
     """
     assert model.model_type == 'classifier', \
         'Now the interpretation visualize only be supported in classifier!'
     if model.status != 'Normal':
-        raise Exception('The interpretation only can deal with the Normal model')
+        raise Exception(
+            'The interpretation only can deal with the Normal model')
     if not osp.exists(save_dir):
         os.makedirs(save_dir)
-    model.arrange_transforms(
-                transforms=model.test_transforms, mode='test')
+    model.arrange_transforms(transforms=model.test_transforms, mode='test')
     tmp_transforms = copy.deepcopy(model.test_transforms)
     tmp_transforms.transforms = tmp_transforms.transforms[:-2]
     img = tmp_transforms(img_file)[0]
     img = np.around(img).astype('uint8')
     img = np.expand_dims(img, axis=0)
     interpreter = None
-    interpreter = get_lime_interpreter(img, model, num_samples=num_samples, batch_size=batch_size)
+    interpreter = get_lime_interpreter(
+        img, model, num_samples=num_samples, batch_size=batch_size)
     img_name = osp.splitext(osp.split(img_file)[-1])[0]
     interpreter.interpret(img, save_dir=save_dir)
-    
-    
-def normlime(img_file, 
-              model, 
-              dataset=None,
-              num_samples=3000, 
-              batch_size=50,
-              save_dir='./'):
+
+
+def normlime(img_file,
+             model,
+             dataset=None,
+             num_samples=3000,
+             batch_size=50,
+             save_dir='./',
+             normlime_weights_file=None):
     """使用NormLIME算法将模型预测结果的可解释性可视化。
-    
+
     NormLIME是利用一定数量的样本来出一个全局的解释。NormLIME会提前计算一定数量的测
     试样本的LIME结果，然后对相同的特征进行权重的归一化，这样来得到一个全局的输入和输出的关系。
-    
+
     注意1：dataset读取的是一个数据集，该数据集不宜过大，否则计算时间会较长，但应包含所有类别的数据。
     注意2：NormLIME可解释性结果可视化目前只支持分类模型。
-         
+
     Args:
         img_file (str): 预测图像路径。
         model (paddlex.cv.models): paddlex中的模型。
         dataset (paddlex.datasets): 数据集读取器，默认为None。
         num_samples (int): LIME用于学习线性模型的采样数，默认为3000。
         batch_size (int): 预测数据batch大小，默认为50。
-        save_dir (str): 可解释性可视化结果（保存为png格式文件）和中间文件存储路径。        
+        save_dir (str): 可解释性可视化结果（保存为png格式文件）和中间文件存储路径。
+        normlime_weights_file (str): NormLIME初始化文件名，若不存在，则计算一次，保存于该路径；若存在，则直接载入。
     """
     assert model.model_type == 'classifier', \
         'Now the interpretation visualize only be supported in classifier!'
     if model.status != 'Normal':
-        raise Exception('The interpretation only can deal with the Normal model')
+        raise Exception(
+            'The interpretation only can deal with the Normal model')
     if not osp.exists(save_dir):
         os.makedirs(save_dir)
-    model.arrange_transforms(
-                transforms=model.test_transforms, mode='test')
+    model.arrange_transforms(transforms=model.test_transforms, mode='test')
     tmp_transforms = copy.deepcopy(model.test_transforms)
     tmp_transforms.transforms = tmp_transforms.transforms[:-2]
     img = tmp_transforms(img_file)[0]
@@ -100,52 +100,48 @@ def normlime(img_file,
     img = np.expand_dims(img, axis=0)
     interpreter = None
     if dataset is None:
-        raise Exception('The dataset is None. Cannot implement this kind of interpretation')
-    interpreter = get_normlime_interpreter(img, model, dataset, 
-                                 num_samples=num_samples, batch_size=batch_size,
-                                     save_dir=save_dir)
+        raise Exception(
+            'The dataset is None. Cannot implement this kind of interpretation')
+    interpreter = get_normlime_interpreter(
+        img,
+        model,
+        dataset,
+        num_samples=num_samples,
+        batch_size=batch_size,
+        save_dir=save_dir,
+        normlime_weights_file=normlime_weights_file)
     img_name = osp.splitext(osp.split(img_file)[-1])[0]
     interpreter.interpret(img, save_dir=save_dir)
-    
-    
+
+
 def get_lime_interpreter(img, model, num_samples=3000, batch_size=50):
     def predict_func(image):
-        image = image.astype('float32')
-        for i in range(image.shape[0]):
-            image[i] = cv2.cvtColor(image[i], cv2.COLOR_RGB2BGR)
-        tmp_transforms = copy.deepcopy(model.test_transforms.transforms)
-        model.test_transforms.transforms = model.test_transforms.transforms[-2:]
         out = interpretation_predict(model, image)
-        model.test_transforms.transforms = tmp_transforms
         return out[0]
+
     labels_name = None
     if hasattr(model, 'labels'):
         labels_name = model.labels
-    interpreter = Interpretation('lime', 
-                            predict_func,
-                            labels_name,
-                            num_samples=num_samples, 
-                            batch_size=batch_size)
+    interpreter = Interpretation(
+        'lime',
+        predict_func,
+        labels_name,
+        num_samples=num_samples,
+        batch_size=batch_size)
     return interpreter
 
 
-def get_normlime_interpreter(img, model, dataset, num_samples=3000, batch_size=50, save_dir='./'):
-    def precompute_predict_func(image):
-        image = image.astype('float32')
-        tmp_transforms = copy.deepcopy(model.test_transforms.transforms)
-        model.test_transforms.transforms = model.test_transforms.transforms[-2:]
-        out = interpretation_predict(model, image)
-        model.test_transforms.transforms = tmp_transforms
-        return out[0]
+def get_normlime_interpreter(img,
+                             model,
+                             dataset,
+                             num_samples=3000,
+                             batch_size=50,
+                             save_dir='./',
+                             normlime_weights_file=None):
     def predict_func(image):
-        image = image.astype('float32')
-        for i in range(image.shape[0]):
-            image[i] = cv2.cvtColor(image[i], cv2.COLOR_RGB2BGR)
-        tmp_transforms = copy.deepcopy(model.test_transforms.transforms)
-        model.test_transforms.transforms = model.test_transforms.transforms[-2:]
         out = interpretation_predict(model, image)
-        model.test_transforms.transforms = tmp_transforms
         return out[0]
+
     labels_name = None
     if dataset is not None:
         labels_name = dataset.labels
@@ -157,28 +153,29 @@ def get_normlime_interpreter(img, model, dataset, num_samples=3000, batch_size=5
             os.makedirs(root_path)
         url = "https://bj.bcebos.com/paddlex/interpret/pre_models.tar.gz"
         pdx.utils.download_and_decompress(url, path=root_path)
-    npy_dir = precompute_for_normlime(precompute_predict_func, 
-                                      dataset, 
-                                      num_samples=num_samples, 
-                                      batch_size=batch_size,
-                                      save_dir=save_dir)
-    interpreter = Interpretation('normlime', 
-                            predict_func,
-                            labels_name,
-                            num_samples=num_samples, 
-                            batch_size=batch_size,
-                            normlime_weights=npy_dir)
-    return interpreter
-
 
-def precompute_for_normlime(predict_func, dataset, num_samples=3000, batch_size=50, save_dir='./'):
-    image_list = []
-    for item in dataset.file_list:
-        image_list.append(item[0])
-    return precompute_normlime_weights(
-            image_list,  
+    if osp.exists(osp.join(save_dir, normlime_weights_file)):
+        normlime_weights_file = osp.join(save_dir, normlime_weights_file)
+        try:
+            np.load(normlime_weights_file, allow_pickle=True).item()
+        except:
+            normlime_weights_file = precompute_global_classifier(
+                dataset,
+                predict_func,
+                save_path=normlime_weights_file,
+                batch_size=batch_size)
+    else:
+        normlime_weights_file = precompute_global_classifier(
+            dataset,
             predict_func,
-            num_samples=num_samples, 
-            batch_size=batch_size,
-            save_dir=save_dir)
-  
+            save_path=normlime_weights_file,
+            batch_size=batch_size)
+
+    interpreter = Interpretation(
+        'normlime',
+        predict_func,
+        labels_name,
+        num_samples=num_samples,
+        batch_size=batch_size,
+        normlime_weights=normlime_weights_file)
+    return interpreter

tutorials/interpret/normlime.py

@@ -14,18 +14,33 @@ model_file = 'https://bj.bcebos.com/paddlex/interpret/mini_imagenet_veg_mobilene
 pdx.utils.download_and_decompress(model_file, path='./')
 
 # 加载模型
-model = pdx.load_model('mini_imagenet_veg_mobilenetv2')
+model_file = 'mini_imagenet_veg_mobilenetv2'
+model = pdx.load_model(model_file)
 
 # 定义测试所用的数据集
+dataset = 'mini_imagenet_veg'
 test_dataset = pdx.datasets.ImageNet(
-    data_dir='mini_imagenet_veg',
-    file_list=osp.join('mini_imagenet_veg', 'test_list.txt'),
-    label_list=osp.join('mini_imagenet_veg', 'labels.txt'),
+    data_dir=dataset,
+    file_list=osp.join(dataset, 'test_list.txt'),
+    label_list=osp.join(dataset, 'labels.txt'),
     transforms=model.test_transforms)
 
-# 可解释性可视化
-pdx.interpret.normlime(
-         'mini_imagenet_veg/mushroom/n07734744_1106.JPEG', 
-          model, 
-          test_dataset, 
-          save_dir='./')
+import numpy as np
+np.random.seed(5)
+perm = np.random.permutation(len(test_dataset.file_list))
+
+for i in range(len(test_dataset.file_list)):
+
+    # 可解释性可视化
+    pdx.interpret.normlime(
+        test_dataset.file_list[perm[i]][0],
+        model,
+        test_dataset,
+        save_dir='./',
+        normlime_weights_file='{}_{}.npy'.format(
+            dataset.split('/')[-1], model.model_name))
+
+    if i == 1:
+        # first iter will have an initialization process, followed by the interpretation.
+        # second iter will directly load the initialization process, followed by the interpretation.
+        break