fix: add small corrections

antecedent/image.py

@@ -94,11 +94,11 @@ class Image:
         self.trivial_search(blocks)
         n = len(blocks)
         for i, block in enumerate(blocks):
-            self.init_shared_dict(shared_dict, task_id, i, n)
-            if block.status[self.pipeline] == 0:
+            self.update_shared_dict(shared_dict, task_id, i, n)
+            if self.pipeline not in block.status:
                 block.search_antecedent(self.pipeline, max_iter, shared_dict, task_id, verbose)
 
-    def init_shared_dict(self, shared_dict, task_id, i, n, completed=0, total=0):
+    def update_shared_dict(self, shared_dict, task_id, i, n, completed=0, total=0):
         if shared_dict is not None and task_id is not None:
             shared_dict[task_id] = {"filename": self.filename,
                                     "current_block": i + 1,
@@ -119,7 +119,7 @@ class Image:
         all_status = []
         n = len(single_channel_blocks)
         for i, block in enumerate(single_channel_blocks):
-            self.init_shared_dict(shared_dict, task_id, i, n)
+            self.update_shared_dict(shared_dict, task_id, i, n)
             if block.status[grayscale_pipeline] == 1:  # Solved with the trivial search
                 antecedents.append(block.antecedents[grayscale_pipeline])
                 all_status.append(1)
@@ -148,14 +148,17 @@ class Image:
             block.iterations[self.pipeline] = np.sum(iterations)
             block.status[self.pipeline] = status
 
-    def trivial_search(self, blocks, pipeline=None):
+    def trivial_search(self, blocks, shared_dict=None, task_id=None, pipeline=None):
         if pipeline is None:
             pipeline = self.pipeline
-        block_values = np.concatenate([block.value for block in blocks])
+        block_values = np.stack([block.value.reshape(-1,8,8) for block in blocks])
         starts = pipeline.backward(block_values)
         transformed_starts = pipeline.forward(starts)
+        completed = 0
         for i in range(len(starts)):
             if np.allclose(block_values[i], transformed_starts[i]):
+                completed += 1
+                self.update_shared_dict(shared_dict, task_id, i, len(starts), completed=completed, total=len(starts))
                 blocks[i].antecedents[pipeline] = starts[i]
                 blocks[i].iterations[pipeline] = 0
                 blocks[i].status[pipeline] = 1
@@ -175,7 +178,7 @@ class Image:
                              'NodefileStart': parameters['node_file_start'],
                              'Cutoff': parameters['cutoff']}
         for i, block in enumerate(blocks):
-            self.init_shared_dict(shared_dict, task_id, i, len(blocks), 0, 0)
+            self.update_shared_dict(shared_dict, task_id, i, len(blocks), 0, 0)
             block.search_antecedent_ilp(gurobi_parameters, shared_dict, task_id, verbose)
 
     def classify(self, likelihood_file):

antecedent/run_on_dataset.py

@@ -123,7 +123,7 @@ def to_csv(path, image):
                 status = block.status[pipeline]
                 antecedent = block.antecedents[pipeline]
                 iteration_heuristic = block.iterations[pipeline]
-                if status != 1 and block.is_clipped[pipeline]:
+                if status != 1 and (pipeline in block.is_clipped and block.is_clipped[pipeline]):
                     status = 3
                     antecedent = None
                     iteration_heuristic = 0
@@ -134,7 +134,7 @@ def to_csv(path, image):
                     sol = None
 
                 writer.writerow(
-                    [filename, str(block.pos), str(pipeline), str(status), str(iteration_heuristic),
+                    [filename, str(pos), str(pipeline), str(status), str(iteration_heuristic),
                      str(iteration_gurobi), sol])
 
 
@@ -147,7 +147,8 @@ def validate_config(config):
         raise ValueError(f"starting_index must be lower than ending_index.")
     if data['preprocessing']['percentage_block_per_image'] <= 0:
         raise ValueError("The percentage of block used per image must be strictly positive")
-    create_pipeline(antecedent_search['pipeline'], antecedent_search['quality'], True)
+    # TODO: improve pipeline verification
+    # create_pipeline(antecedent_search['pipeline'], antecedent_search['quality'], True)
 
 
 def create_output_folder(config, config_path, verbose):