标准化数值

Question

如何标准化 num_of_lanes 值，最小车道数为 1，最大车道数为 6，我能够创建数据，但如何才能标准化 num_of_lanes，而不是根据其值将它们从 1 到 6 进行标准化。我希望显示保存在 .pkl 和 .txt 文件中的值，但改为标准化值。

这是我的代码

import os
import xml.etree.ElementTree as ET
import pickle

# Define the Directories
annotations_dir = 
attributes_dir = 
output_dir =

# Create subdirectories for pkl and txt within the output directory
output_dir_pkl = os.path.join(output_dir, 'pkl')
output_dir_txt = os.path.join(output_dir, 'txt')

# Check if output subfolders exist, if not, create them
if not os.path.exists(output_dir_pkl):
    os.makedirs(output_dir_pkl)
if not os.path.exists(output_dir_txt):
    os.makedirs(output_dir_txt)

# Get a sorted list of XML files to maintain order in the annotations directory
xml_files = sorted([f for f in os.listdir(annotations_dir) if f.endswith('.xml')])

# Get a sorted list of attribute XML files to maintain order
num_of_lanes_files = sorted([f for f in os.listdir(attributes_dir) if f.endswith('_attributes.xml')])

# Parse the num_of_lanes information from the attribute XML files
num_of_lanes_mapping = {}
default_num_of_lanes_per_video = {}

for num_of_lanes_file in num_of_lanes_files:
    num_of_lanes_tree = ET.parse(os.path.join(attributes_dir, num_of_lanes_file))
    num_of_lanes_root = num_of_lanes_tree.getroot()
    video_name = num_of_lanes_file.split('_attributes')[0]

    default_num_of_lanes = None
    for pedestrian in num_of_lanes_root.findall('.//pedestrian'):
        ped_id = pedestrian.get('id')
        old_id = pedestrian.get('old_id')
        num_of_lanes = pedestrian.get('num_lanes')
        num_of_lanes_mapping[(video_name, ped_id, old_id)] = num_of_lanes

        if default_num_of_lanes is None:
            default_num_of_lanes = num_of_lanes

    default_num_of_lanes_per_video[video_name] = default_num_of_lanes

# Process main XML files
for filename in xml_files:
    file_path = os.path.join(annotations_dir, filename)
    tree = ET.parse(file_path)
    root = tree.getroot()
    video_name = root.find('.//name').text if root.find('.//name') is not None else filename.split('.xml')[0]

    num_of_lanes_data = {}

    for track in root.findall('.//track'):
        for box in track.findall('box'):
            frame_num = str(box.get('frame')).zfill(5)
            ped_id = box.find("./attribute[@name='id']").text
            old_id = box.find("./attribute[@name='old_id']").text

            key = (video_name, ped_id, old_id)
            ped_num_of_lanes = num_of_lanes_mapping.get(key, default_num_of_lanes_per_video[video_name])

            frame_id = f"{video_name}_{frame_num}_{ped_id}"
            num_of_lanes_data[frame_id] = {'num_lanes': ped_num_of_lanes}

  # Save data in pkl and txt files
    pkl_filename = f"{video_name}_num_of_lanes_data.pkl"
    with open(os.path.join(output_dir_pkl, pkl_filename), 'wb') as pkl_file:
        pickle.dump(num_of_lanes_data, pkl_file)

    txt_filename = f"{video_name}_num_of_lanes_data.txt"
    with open(os.path.join(output_dir_txt, txt_filename), 'w') as txt_file:
        for frame_id, data in num_of_lanes_data.items():
            txt_file.write(f"{frame_id}: num_of_lanes: {data['num_lanes']}\n")

Answer 1

要标准化

num_of_lanes

值，您可以使用以下公式：

归一化值 = 值-最小值/(最大值-最小值)

import os
import xml.etree.ElementTree as ET
import pickle

# Define the Directories
annotations_dir = 
attributes_dir = 
output_dir =

# Define min and max values for normalization
MIN_NUM_LANES = 1
MAX_NUM_LANES = 6

# Check if output subfolders exist, if not, create them
output_dir_pkl = os.path.join(output_dir, 'pkl')
output_dir_txt = os.path.join(output_dir, 'txt')

# Ensure the output subdirectories exist
os.makedirs(output_dir_pkl, exist_ok=True)
os.makedirs(output_dir_txt, exist_ok=True)

# Get a sorted list of attribute XML files to maintain order
num_of_lanes_files = sorted([f for f in os.listdir(attributes_dir) if f.endswith('_attributes.xml')])

for num_of_lanes_file in num_of_lanes_files:
    num_of_lanes_tree = ET.parse(os.path.join(attributes_dir, num_of_lanes_file))
    num_of_lanes_root = num_of_lanes_tree.getroot()
    video_name = num_of_lanes_file.split('_attributes')[0]

    # Assuming you extract num_of_lanes value here; Example:
    # num_of_lanes = int(num_of_lanes_root.find('SomeXPath').text)
    # For demonstration, let's assume num_of_lanes is directly available

    normalized_num_of_lanes = {}  # Dictionary to store normalized values

    for child in num_of_lanes_root:
        # Example of extracting num_of_lanes, replace with actual extraction logic
        num_of_lanes = int(child.text)  # Example placeholder
        normalized_value = (num_of_lanes - MIN_NUM_LANES) / (MAX_NUM_LANES - MIN_NUM_LANES)
        normalized_num_of_lanes[child.tag] = normalized_value

    # Save normalized values to PKL
    pkl_filename = os.path.join(output_dir_pkl, f"{video_name}.pkl")
    with open(pkl_filename, 'wb') as pkl_file:
        pickle.dump(normalized_num_of_lanes, pkl_file)

    # Save normalized values to TXT
    txt_filename = os.path.join(output_dir_txt, f"{video_name}.txt")
    with open(txt_filename, 'w') as txt_file:
        for key, value in normalized_num_of_lanes.items():
            txt_file.write(f"{key}: {value}\n")

print("Normalization and saving completed.")