modified based on upl app.py commit

app.py

@@ -22,7 +22,7 @@ def inference(filename):
     image_height, image_width = image.shape[:2]
     boxes, confs, classes = detector.detect(image)
 
-    for box, conf, class_ind in zip(boxes, confs, classes):
+    for box, _, _ in zip(boxes, confs, classes):
         box = box.astype(np.int32)
         box_width = box[2] - box[0] + 1
         box_height = box[3] - box[1] + 1
@@ -38,7 +38,8 @@ def inference(filename):
         if prob < 0.10:
             text = 'Unknown'
         else:
-            text = '{}: {:.2f}%'.format(
+            text = '{}({}): {:.2f}%'.format(
+                results[0]['chinese_name'],
                 results[0]['latin_name'],
                 100.0 * results[0]['probability']
             )

khandy/label/detect.py

@@ -13,7 +13,7 @@ import lxml.builder
 import numpy as np
 
 
-__all__ = ['DetectIrObject', 'DetectIrRecord', 'load_detect', 
+__all__ = ['DetectIrObject', 'DetectIrRecord', 'load_detect',
            'save_detect', 'convert_detect', 'replace_detect_label',
            'load_coco_class_names']
 
@@ -27,8 +27,8 @@ class DetectIrObject:
     y_min: float
     x_max: float
     y_max: float
-    
-    
+
+
 @dataclass
 class DetectIrRecord:
     """Intermediate Representation Format of Record
@@ -37,30 +37,30 @@ class DetectIrRecord:
     width: int
     height: int
     objects: List[DetectIrObject] = field(default_factory=list)
-    
-    
+
+
 @dataclass
 class PascalVocSource:
     database: str = ''
     annotation: str = ''
     image: str = ''
-    
-    
+
+
 @dataclass
 class PascalVocSize:
     height: int
     width: int
     depth: int
-    
-    
+
+
 @dataclass
 class PascalVocBndbox:
     xmin: float
     ymin: float
     xmax: float
     ymax: float
-    
-    
+
+
 @dataclass
 class PascalVocObject:
     name: str
@@ -68,8 +68,8 @@ class PascalVocObject:
     truncated: int = 0
     difficult: int = 0
     bndbox: Optional[PascalVocBndbox] = None
-    
-    
+
+
 @dataclass
 class PascalVocRecord:
     folder: str = ''
@@ -79,33 +79,33 @@ class PascalVocRecord:
     size: Optional[PascalVocSize] = None
     segmented: int = 0
     objects: List[PascalVocObject] = field(default_factory=list)
-    
-    
+
+
 class PascalVocHandler:
     @staticmethod
     def load(filename, **kwargs) -> PascalVocRecord:
         pascal_voc_record = PascalVocRecord()
-        
+
         xml_tree = ET.parse(filename)
         pascal_voc_record.folder = xml_tree.find('folder').text
         pascal_voc_record.filename = xml_tree.find('filename').text
         pascal_voc_record.path = xml_tree.find('path').text
         pascal_voc_record.segmented = xml_tree.find('segmented').text
-        
+
         source_tag = xml_tree.find('source')
         pascal_voc_record.source = PascalVocSource(
             database=source_tag.find('database').text,
             # annotation=source_tag.find('annotation').text,
             # image=source_tag.find('image').text
         )
-        
+
         size_tag = xml_tree.find('size')
         pascal_voc_record.size = PascalVocSize(
             width=int(size_tag.find('width').text),
             height=int(size_tag.find('height').text),
             depth=int(size_tag.find('depth').text)
         )
-        
+
         object_tags = xml_tree.findall('object')
         for index, object_tag in enumerate(object_tags):
             bndbox_tag = object_tag.find('bndbox')
@@ -124,7 +124,7 @@ class PascalVocHandler:
             )
             pascal_voc_record.objects.append(pascal_voc_object)
         return pascal_voc_record
-        
+
     @staticmethod
     def save(filename, pascal_voc_record: PascalVocRecord):
         maker = lxml.builder.ElementMaker()
@@ -135,14 +135,14 @@ class PascalVocHandler:
             maker.source(
                 maker.database(pascal_voc_record.source.database),
             ),
-            maker.size( 
+            maker.size(
                 maker.width(str(pascal_voc_record.size.width)),
                 maker.height(str(pascal_voc_record.size.height)),
                 maker.depth(str(pascal_voc_record.size.depth)),
             ),
             maker.segmented(str(pascal_voc_record.segmented)),
         )
-        
+
         for pascal_voc_object in pascal_voc_record.objects:
             object_tag = maker.object(
                 maker.name(pascal_voc_object.name),
@@ -157,12 +157,13 @@ class PascalVocHandler:
                 ),
             )
             xml.append(object_tag)
-            
+
         if not filename.endswith('.xml'):
             filename = filename + '.xml'
         with open(filename, 'wb') as f:
-            f.write(lxml.etree.tostring(xml, pretty_print=True, encoding='utf-8'))
-            
+            f.write(lxml.etree.tostring(
+                xml, pretty_print=True, encoding='utf-8'))
+
     @staticmethod
     def to_ir(pascal_voc_record: PascalVocRecord) -> DetectIrRecord:
         ir_record = DetectIrRecord(
@@ -180,7 +181,7 @@ class PascalVocHandler:
             )
             ir_record.objects.append(ir_object)
         return ir_record
-        
+
     @staticmethod
     def from_ir(ir_record: DetectIrRecord) -> PascalVocRecord:
         pascal_voc_record = PascalVocRecord(
@@ -203,10 +204,11 @@ class PascalVocHandler:
             )
             pascal_voc_record.objects.append(pascal_voc_object)
         return pascal_voc_record
-        
-        
+
+
 class _NumpyEncoder(json.JSONEncoder):
     """ Special json encoder for numpy types """
+
     def default(self, obj):
         if isinstance(obj, (np.bool_,)):
             return bool(obj)
@@ -279,7 +281,7 @@ class LabelmeHandler:
             )
             ir_record.objects.append(ir_object)
         return ir_record
-        
+
     @staticmethod
     def from_ir(ir_record: DetectIrRecord) -> LabelmeRecord:
         labelme_record = LabelmeRecord(
@@ -291,12 +293,12 @@ class LabelmeHandler:
             labelme_shape = LabelmeShape(
                 label=ir_object.label,
                 shape_type='rectangle',
-                points=[[ir_object.x_min, ir_object.y_min], 
+                points=[[ir_object.x_min, ir_object.y_min],
                         [ir_object.x_max, ir_object.y_max]]
             )
             labelme_record.shapes.append(labelme_shape)
         return labelme_record
-        
+
 
 @dataclass
 class YoloObject:
@@ -305,16 +307,16 @@ class YoloObject:
     y_center: float
     width: float
     height: float
-    
-    
+
+
 @dataclass
 class YoloRecord:
     filename: Optional[str] = None
     width: Optional[int] = None
     height: Optional[int] = None
     objects: List[YoloObject] = field(default_factory=list)
-    
-    
+
+
 class YoloHandler:
     @staticmethod
     def load(filename, **kwargs) -> YoloRecord:
@@ -341,7 +343,8 @@ class YoloHandler:
     def save(filename, yolo_record: YoloRecord):
         records = []
         for object in yolo_record.objects:
-            records.append(f'{object.label} {object.x_center} {object.y_center} {object.width} {object.height}')
+            records.append(
+                f'{object.label} {object.x_center} {object.y_center} {object.width} {object.height}')
         if not filename.endswith('.txt'):
             filename = filename + '.txt'
         khandy.save_list(filename, records)
@@ -354,10 +357,14 @@ class YoloHandler:
             height=yolo_record.height
         )
         for yolo_object in yolo_record.objects:
-            x_min = (yolo_object.x_center - 0.5 * yolo_object.width) * yolo_record.width
-            y_min = (yolo_object.y_center - 0.5 * yolo_object.height) * yolo_record.height
-            x_max = (yolo_object.x_center + 0.5 * yolo_object.width) * yolo_record.width
-            y_max = (yolo_object.y_center + 0.5 * yolo_object.height) * yolo_record.height
+            x_min = (yolo_object.x_center - 0.5 *
+                     yolo_object.width) * yolo_record.width
+            y_min = (yolo_object.y_center - 0.5 *
+                     yolo_object.height) * yolo_record.height
+            x_max = (yolo_object.x_center + 0.5 *
+                     yolo_object.width) * yolo_record.width
+            y_max = (yolo_object.y_center + 0.5 *
+                     yolo_object.height) * yolo_record.height
             ir_object = DetectIrObject(
                 label=yolo_object.label,
                 x_min=x_min,
@@ -367,7 +374,7 @@ class YoloHandler:
             )
             ir_record.objects.append(ir_object)
         return ir_record
-        
+
     @staticmethod
     def from_ir(ir_record: DetectIrRecord) -> YoloRecord:
         yolo_record = YoloRecord(
@@ -376,8 +383,10 @@ class YoloHandler:
             height=ir_record.height
         )
         for ir_object in ir_record.objects:
-            x_center = (ir_object.x_max + ir_object.x_min) / (2 * ir_record.width)
-            y_center = (ir_object.y_max + ir_object.y_min) / (2 * ir_record.height)
+            x_center = (ir_object.x_max + ir_object.x_min) / \
+                (2 * ir_record.width)
+            y_center = (ir_object.y_max + ir_object.y_min) / \
+                (2 * ir_record.height)
             width = abs(ir_object.x_max - ir_object.x_min) / ir_record.width
             height = abs(ir_object.y_max - ir_object.y_min) / ir_record.height
             yolo_object = YoloObject(
@@ -389,8 +398,8 @@ class YoloHandler:
             )
             yolo_record.objects.append(yolo_object)
         return yolo_record
-    
-        
+
+
 @dataclass
 class CocoObject:
     label: str
@@ -398,29 +407,29 @@ class CocoObject:
     y_min: float
     width: float
     height: float
-    
-    
+
+
 @dataclass
 class CocoRecord:
     filename: str
     width: int
     height: int
     objects: List[CocoObject] = field(default_factory=list)
-    
+
 
 class CocoHandler:
     @staticmethod
     def load(filename, **kwargs) -> List[CocoRecord]:
         json_data = khandy.load_json(filename)
-        
+
         images = json_data['images']
         annotations = json_data['annotations']
         categories = json_data['categories']
-        
+
         label_map = {}
         for cat_item in categories:
             label_map[cat_item['id']] = cat_item['name']
-        
+
         coco_records = OrderedDict()
         for image_item in images:
             coco_records[image_item['id']] = CocoRecord(
@@ -428,7 +437,7 @@ class CocoHandler:
                 width=image_item['width'],
                 height=image_item['height'],
                 objects=[])
-                
+
         for annotation_item in annotations:
             coco_object = CocoObject(
                 label=label_map[annotation_item['category_id']],
@@ -436,9 +445,10 @@ class CocoHandler:
                 y_min=annotation_item['bbox'][1],
                 width=annotation_item['bbox'][2],
                 height=annotation_item['bbox'][3])
-            coco_records[annotation_item['image_id']].objects.append(coco_object)
+            coco_records[annotation_item['image_id']
+                         ].objects.append(coco_object)
         return list(coco_records.values())
-        
+
     @staticmethod
     def to_ir(coco_record: CocoRecord) -> DetectIrRecord:
         ir_record = DetectIrRecord(
@@ -474,8 +484,8 @@ class CocoHandler:
             )
             coco_record.objects.append(coco_object)
         return coco_record
-        
-        
+
+
 def load_detect(filename, fmt, **kwargs) -> DetectIrRecord:
     if fmt == 'labelme':
         labelme_record = LabelmeHandler.load(filename, **kwargs)
@@ -488,12 +498,13 @@ def load_detect(filename, fmt, **kwargs) -> DetectIrRecord:
         ir_record = PascalVocHandler.to_ir(pascal_voc_record)
     elif fmt == 'coco':
         coco_records = CocoHandler.load(filename, **kwargs)
-        ir_record = [CocoHandler.to_ir(coco_record) for coco_record in coco_records]
+        ir_record = [CocoHandler.to_ir(coco_record)
+                     for coco_record in coco_records]
     else:
         raise ValueError(f"Unsupported detect label fmt. Got {fmt}")
     return ir_record
-    
-    
+
+
 def save_detect(filename, ir_record: DetectIrRecord, out_fmt):
     os.makedirs(os.path.dirname(os.path.abspath(filename)), exist_ok=True)
     if out_fmt == 'labelme':
@@ -527,9 +538,11 @@ def _get_format(record):
 
 
 def convert_detect(record, out_fmt):
-    allowed_fmts = ('labelme', 'yolo', 'voc', 'coco', 'pascal', 'pascal_voc', 'ir', 'detect_ir')
+    allowed_fmts = ('labelme', 'yolo', 'voc', 'coco',
+                    'pascal', 'pascal_voc', 'ir', 'detect_ir')
     if out_fmt not in allowed_fmts:
-        raise ValueError("Unsupported label format conversions for given out_fmt")
+        raise ValueError(
+            "Unsupported label format conversions for given out_fmt")
     if out_fmt in _get_format(record):
         return record
 
@@ -545,7 +558,7 @@ def convert_detect(record, out_fmt):
         ir_record = record
     else:
         raise TypeError('Unsupported type for record')
-        
+
     if out_fmt in ('ir', 'detect_ir'):
         dst_record = ir_record
     elif out_fmt == 'labelme':
@@ -557,7 +570,7 @@ def convert_detect(record, out_fmt):
     elif out_fmt == 'coco':
         dst_record = CocoHandler.from_ir(ir_record)
     return dst_record
-    
+
 
 def replace_detect_label(record: DetectIrRecord, label_map, ignore=True):
     dst_record = copy.deepcopy(record)
@@ -579,4 +592,3 @@ def load_coco_class_names(filename):
     json_data = khandy.load_json(filename)
     categories = json_data['categories']
     return [cat_item['name'] for cat_item in categories]
-

khandy/split_utils.py

@@ -1,5 +1,5 @@
 import numbers
-from collections import Sequence
+from collections.abc import Sequence
 
 import numpy as np
 
@@ -15,7 +15,7 @@ def split_by_num(x, num_splits, strict=True):
     # NB: np.ndarray is not Sequence
     assert isinstance(x, (Sequence, np.ndarray))
     assert isinstance(num_splits, numbers.Integral)
-    
+
     if strict:
         assert len(x) % num_splits == 0
     split_size = (len(x) + num_splits - 1) // num_splits
@@ -23,8 +23,8 @@ def split_by_num(x, num_splits, strict=True):
     for i in range(0, len(x), split_size):
         out_list.append(x[i: i + split_size])
     return out_list
-    
-    
+
+
 def split_by_size(x, sizes):
     """
     References:
@@ -34,7 +34,7 @@ def split_by_size(x, sizes):
     # NB: np.ndarray is not Sequence
     assert isinstance(x, (Sequence, np.ndarray))
     assert isinstance(sizes, (list, tuple))
-     
+
     assert sum(sizes) == len(x)
     out_list = []
     start_index = 0
@@ -42,8 +42,8 @@ def split_by_size(x, sizes):
         out_list.append(x[start_index: start_index + size])
         start_index += size
     return out_list
-    
-    
+
+
 def split_by_slice(x, slices):
     """
     References:
@@ -52,7 +52,7 @@ def split_by_slice(x, slices):
     # NB: np.ndarray is not Sequence
     assert isinstance(x, (Sequence, np.ndarray))
     assert isinstance(slices, (list, tuple))
-    
+
     out_list = []
     indices = [0] + list(slices) + [len(x)]
     for i in range(len(slices) + 1):
@@ -64,10 +64,8 @@ def split_by_ratio(x, ratios):
     # NB: np.ndarray is not Sequence
     assert isinstance(x, (Sequence, np.ndarray))
     assert isinstance(ratios, (list, tuple))
-    
+
     pdf = [k / sum(ratios) for k in ratios]
     cdf = [sum(pdf[:k]) for k in range(len(pdf) + 1)]
     indices = [int(round(len(x) * k)) for k in cdf]
     return [x[indices[i]: indices[i + 1]] for i in range(len(ratios))]
-    
-