Python - 分配错误之前引用的局部变量

Question

I am trying to adopt mask_rcnn original code to my dataset.The original code takes xml files and I just modified my code to accept json format. Here is the code

   def load_mask(self, image_id, xml=False):
        # get details of image
        info = self.image_info[image_id]

        if xml:
            # define anntation  file location
            path = info['annotation']
            # load XML
            boxes, w, h = self.extract_boxes(path, xml=True)
        else:
            with open('D:\Mask_RCNN\Mask_RCNN\dataset\\annots\\annotations.json', 'r') as f:
                annots = json.load(f)
                found = False
                for value in annots.values():
                    if 'image' in value and value['instance_list']:
                        if value['image']['original_filename'][:-3] == info['id']:
                            boxes, w, h = self.extract_boxes(value)
                            found = True
                            break

        if found:
            # create one array for all masks, each on a different channel
            masks = zeros([h, w, len(boxes)], dtype='uint8')
        else:
            stop = "here"

                if found:
            # create one array for all masks, each on a different channel
            masks = zeros([h, w, len(boxes)], dtype='uint8')
        else:
            stop = "here"

        # create masks
        class_ids = list()
        for i in range(len(boxes)):
            box = boxes[i]
            row_s, row_e = box[1], box[3]
            col_s, col_e = box[0], box[2]
            masks[row_s:row_e, col_s:col_e, i] = 1
            class_ids.append(self.class_names.index('Penquins'))
        return masks, asarray(class_ids, dtype='int32')

    # load an image reference
        #"""Return the path of the image."""
    def image_reference(self, image_id):
            info = self.image_info[image_id]
            print(info)
            return info['path']

It gives me the error

File "C:/PycharmProjects/Mask_RCNN/Mask_RCNN/objects.py", line 197, in load_mask
for i in range(len(boxes)): UnboundLocalError: local variable 'boxes' referenced before assignment

I tried to debug the code,its throwing an error before creating masks but I am not able to figure out whats wrong. Any idea?

Answer 1

Only if xml is true, it is guaranteed, that boxes gets a value. Otherwise additional conditions have to be fulfilled, which does not seem to be the case.

Answer 2

boxes is initialized in some of the scenarios in load_mask 's first if else block, which means it might not exists when you reach if found: . Just declare it in the beginning

def load_mask(self, image_id, xml=False):
    # get details of image
    info = self.image_info[image_id]
    boxes = []

Answer 3

It means that you are trying to iterate over the number of boxes , but your if statements above are constructed in such a way that boxes might not exit by the time this statement gets executed.

try this simple example:

user_input = "no"

if user_input =="yes":
    boxes = [1,2,3,4]
else:
    pass

print(boxes)

This will result in the same error as you are seeing. Your problem is that both of these if statements must pass before boxes is defined:

           if 'image' in value and value['instance_list']:
                if value['image']['original_filename'][:-3] == info['id']:

Likely you either need additional logic that ensures the iteration only occurs if certain conditions are met. Or, if boxes is a list you could start your code by defining boxes=[] . In that case the iteration will simply pass over without doing anything as the length of boxes is zero in that case

Answer 4

The problem is that boxes isnt getting declared in line of this for loop, so define boxes as an empty list in the first part of the function

def load_mask(self, image_id, xml=False):
    boxes = []
    #do all the other stuff here

Then when you do the for loop it should now be declared (unless one of the functions don't return the boxes and give you something else)

Quick tip when doing the for loop, you can do

for i, box in enumerate(boxes):
    ...

which will not make you define box inside the loop (i will be 0 to len(boxes)-1 and box will be the box with that index)