nonunique error fix + code improvements

src/compute/layers/processing/SplitDataLayer.py

@@ -52,61 +52,45 @@ def modifies_data(self):
         return False
 
     def process(self, data_el: Tuple[ImageDescriptor, Annotation]):
-        def _split_by_percent() -> List[Tuple[ImageDescriptor, Annotation]]:
+        def replace_ds_name(item_desc, new_ds_name):
             new_item_desc = deepcopy(item_desc)
+            new_item_desc.res_ds_name = new_ds_name
+            return new_item_desc
+
+        def _split_by_percent() -> List[Tuple[ImageDescriptor, Annotation]]:
             total_items_cnt = self.net.total_elements_cnt
             split_ratio = self.settings["split_ratio"]
             split_num = total_items_cnt * split_ratio / 100
             split_index = int(item_idx / split_num) + (item_idx % split_num > 0)
             dataset = f"split_{split_index}"
-            new_item_desc.res_ds_name = dataset
-            return [(new_item_desc, ann)]
+            return [(replace_ds_name(item_desc, dataset), ann)]
 
         def _split_by_num() -> List[Tuple[ImageDescriptor, Annotation]]:
-            new_item_desc = deepcopy(item_desc)
             split_num = self.settings["split_num"]
             split_index = int(item_idx / split_num) + (item_idx % split_num > 0)
             dataset = f"split_{split_index}"
-            new_item_desc.res_ds_name = dataset
-            return [(new_item_desc, ann)]
+            return [(replace_ds_name(item_desc, dataset), ann)]
 
         def _split_by_class() -> List[Tuple[ImageDescriptor, Annotation]]:
             image_labels = ann.labels
             if len(image_labels) > 0:
                 classes = list({label.obj_class.name for label in image_labels})
-                items = []
-                for class_name in classes:
-                    curr_item_desc = deepcopy(item_desc)
-                    curr_item_desc.res_ds_name = class_name
-                    items.append((curr_item_desc, ann))
-                return items
+                return [(replace_ds_name(item_desc, class_name), ann) for class_name in classes]
             else:
-                new_item_desc = deepcopy(item_desc)
-                new_item_desc.res_ds_name = "unlabeled"
-                return [(new_item_desc, ann)]
+                return [(replace_ds_name(item_desc, "unlabeled"), ann)]
 
         def _split_by_tags() -> List[Tuple[ImageDescriptor, Annotation]]:
             image_tags = list(set(ann.img_tags.keys()))
             label_tags = list(set([tag for label in ann.labels for tag in label.tags.keys()]))
             if len(image_tags) == 0 and len(label_tags) == 0:
-                new_item_desc = deepcopy(item_desc)
-                new_item_desc.res_ds_name = "no tags"
-                return [(new_item_desc, ann)]
+                return [(replace_ds_name(item_desc, "no tags"), ann)]
 
             tag_names = set()
             items = []
-            for img_tag_name in image_tags:
-                if img_tag_name not in tag_names:
-                    tag_names.add(img_tag_name)
-                    new_img_item_desc = deepcopy(item_desc)
-                    new_img_item_desc.res_ds_name = img_tag_name
-                    items.append((new_img_item_desc, ann))
-            for tag_name in label_tags:
-                if tag_name not in tag_names:
-                    tag_names.add(tag_name)
-                    new_label_item_desc = deepcopy(item_desc)
-                    new_label_item_desc.res_ds_name = tag_name
-                    items.append((new_label_item_desc, ann))
+            for tag in image_tags + label_tags:
+                if tag not in tag_names:
+                    tag_names.add(tag)
+                    items.append((replace_ds_name(item_desc, tag), ann))
             return items
 
         if self.net.preview_mode:

src/compute/layers/save/CreateNewProjectLayer.py

@@ -131,18 +131,19 @@ def process_batch(self, data_els: List[Tuple[ImageDescriptor, Annotation]]):
                 ds_item_map[dataset_name].append((item_desc, ann))
 
             for ds_name in ds_item_map:
-                out_item_names = [
-                    self.get_free_name(
-                        item_desc.get_item_name(), dataset_name, self.out_project_name
-                    )
-                    + get_file_ext(item_desc.info.item_info.name)
-                    for item_desc, _ in ds_item_map[ds_name]
-                ]
                 if self.sly_project_info is not None:
                     # @TODO: not safe, fix later
                     orig_ds_info = ds_item_map[ds_name][0][0].info.ds_info
                     ds_parents = self.get_ds_parents(orig_ds_info)
                     dataset_info = self.get_or_create_dataset(ds_name, ds_parents)
+
+                    out_item_names = [
+                        self.get_free_name(
+                            item_desc.get_item_name(), ds_name, self.out_project_name
+                        )
+                        + get_file_ext(item_desc.info.item_info.name)
+                        for item_desc, _ in ds_item_map[ds_name]
+                    ]
                     if self.net.modality == "images":
                         if self.net.may_require_items():
                             item_infos = g.api.image.upload_nps(
@@ -151,14 +152,17 @@ def process_batch(self, data_els: List[Tuple[ImageDescriptor, Annotation]]):
                                 [item_desc.read_image() for item_desc, _ in ds_item_map[ds_name]],
                             )
                         else:
-                            item_infos = g.api.image.upload_ids(
-                                dataset_info.id,
-                                out_item_names,
-                                [
-                                    item_desc.info.item_info.id
-                                    for item_desc, _ in ds_item_map[ds_name]
-                                ],
-                            )
+                            try:
+                                item_infos = g.api.image.upload_ids(
+                                    dataset_info.id,
+                                    out_item_names,
+                                    [
+                                        item_desc.info.item_info.id
+                                        for item_desc, _ in ds_item_map[ds_name]
+                                    ],
+                                )
+                            except:
+                                pass
                         g.api.annotation.upload_anns(
                             [item_info.id for item_info in item_infos],
                             [ann for _, ann in ds_item_map[ds_name]],