pybaseutils 2.4.35__tar.gz → 2.5.2__tar.gz

{pybaseutils-2.4.35 → pybaseutils-2.5.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pybaseutils
-Version: 2.4.35
+Version: 2.5.2
 Summary: pybaseutils
 Home-page: https://github.com/PanJinquan/base-utils
 Author: PanJinquan

{pybaseutils-2.4.35 → pybaseutils-2.5.2}/pybaseutils/__init__.py

@@ -8,4 +8,4 @@
 # --------------------------------------------------------
 """
 
-__version__ = '2.4.35'
+__version__ = '2.5.2'

{pybaseutils-2.4.35 → pybaseutils-2.5.2}/pybaseutils/audio/audio_utils.py

@@ -19,7 +19,6 @@ from ffmpy import FFmpeg
 import matplotlib.pyplot as plt
 from scipy.fft import fft
 from IPython.display import Audio, display
-from playsound import playsound
 from scipy.io import wavfile
 
 import numpy as np
@@ -267,6 +266,8 @@ def sound_audio(audio, sr=16000):
 
 def sound_audio_file(audio_file, jupyter=False):
     """
+    sudo apt-get install libasound2-dev
+    pip install simpleaudio==1.0.4
     :param audio_file:
     :param jupyter:
     :return:
@@ -274,7 +275,10 @@ def sound_audio_file(audio_file, jupyter=False):
     if jupyter:
         display(Audio(audio_file))
     else:
-        playsound(audio_file)
+        import simpleaudio as sa
+        wave_obj = sa.WaveObject.from_wave_file(audio_file)
+        play_obj = wave_obj.play() # 播放
+        play_obj.wait_done()  # 等待播放结束（可选，但推荐）
 
 
 def sound_audio_buffer(audio_buffer, tmp_audio="./tmp.wav"):
@@ -470,8 +474,8 @@ if __name__ == '__main__':
     video_out = "../../data/video/test-video-result.mp4"
     audio_file = "../../data/audio/bus_chinese.wav"
 
-    test_time(audio_file)
-    # playsound(audio_file)
+    # test_time(audio_file)
+    sound_audio_file(audio_file,jupyter=False)
     # playsound(audio_file)
     # display_time_domain(audio_file)
     # display_freq_domain(audio_file)

{pybaseutils-2.4.35 → pybaseutils-2.5.2}/pybaseutils/color_utils.py

@@ -11,7 +11,6 @@ import cv2
 import matplotlib.pyplot as plt
 import numpy as np
 from collections import defaultdict, Counter
-from pybaseutils import file_utils
 
 VOC_NAMES = ["aeroplane", "bicycle", "bird", "boat", "bottle", "bus", "car", "cat",
              "chair", "cow", "diningtable", "dog", "horse", "motorbike", "person",
@@ -38,7 +37,7 @@ color_table = [(0, 0, 0), (0, 255, 0), (255, 0, 0), (0, 0, 255), (0, 255, 255),
                (0, 64, 0), (128, 64, 0), (0, 192, 0), (128, 192, 0), (0, 64, 128)] * 100
 
 
-class ColorHash():
+class ColorLabel():
     def __init__(self):
         self.color = dict()
         self.cnums = len(color_table)
@@ -48,7 +47,7 @@ class ColorHash():
         :param label: str or list
         :return:
         """
-        h = file_utils.get_hash(texts=[str(label)])
+        h = str(label)
         if h not in self.color:
             i = len(self.color) + 1
             self.color[h] = color_table[i % self.cnums]

{pybaseutils-2.4.35 → pybaseutils-2.5.2}/pybaseutils/converter/build_coco.py

@@ -211,6 +211,7 @@ class COCOBuilder():
                 keypoints[1::3] = np.clip(keypoints[1::3], 1, height - 1)
         :return:
         """
+        if isinstance(keypoints, np.ndarray): keypoints = keypoints.tolist()
         self.annotation_id += 1
         annotation_item = dict()
         annotation_item['segmentation'] = seg
@@ -274,24 +275,21 @@ class COCOBuilder():
         if area < min_area: segs, area = [], []
         return segs, area
 
-    def addObjects(self, filename, objects: dict, width, height, num_joints, ignore_tiny=False):
+    def addObjects(self, filename, objects: dict, width, height, ignore_tiny=False):
         """
         :param filename: os.path.basename(filename)
         :param objects:
-                        boxes  is [[xmin,ymin,xmax,ymax],[xmin,ymin,xmax,ymax],...]
-                        labels is [目标1,目标2,...]
-                        contours is [目标1的轮廓,目标2的轮廓,...]
-                        keypoints is [目标1的关键点,目标2的关键点,...],格式：
-                        keypoints=num_joints*3,x,y,visibility
-                        keypoints关节点的格式 : [x_1, y_1, v_1,...,x_k, y_k, v_k]
-                        其中x,y为Keypoint的坐标，v为可见标志
-                            v = 0 : 未标注点
-                            v = 1 : 标注了但是图像中不可见（例如遮挡）
-                            v = 2 : 标注了并图像可见
+                        boxes     [[xmin,ymin,xmax,ymax],[xmin,ymin,xmax,ymax],...]
+                        labels    [目标1,目标2,...]
+                        contours  [目标1[seg1,seg2,...],目标2[seg1,seg2,...],...]
+                        keypoints [目标1[x1, y1, v1,x2, y2, v2,...],目标1[x1, y1, v1,x2, y2, v2,...],...]
+                                  其中x,y为Keypoint的坐标，v为可见标志
+                                        v = 0 : 未标注点
+                                        v = 1 : 标注了但是图像中不可见（例如遮挡）
+                                        v = 2 : 标注了并图像可见
         实际预测时，不要求预测每个关节点的可见性
         :param width:
         :param height:
-        :param num_joints:
         :param ignore_tiny: 是否去除segment很细小的目标
         :return:
         """

pybaseutils-2.5.2/pybaseutils/converter/convert_labelme2coco.py

@@ -0,0 +1,122 @@
+# -*- coding: utf-8 -*-
+"""
+# --------------------------------------------------------
+# @Project:
+# @Author : panjq
+# @Date   : 2020-02-12 18:28:16
+# @url    :
+# --------------------------------------------------------
+"""
+
+import argparse
+import sys
+import os
+import numpy as np
+import json
+import cv2
+import time
+from tqdm import tqdm
+from pybaseutils.dataloader import parser_labelme
+from pybaseutils.converter import build_coco
+
+
+class Labelme2COCO(build_coco.COCOBuilder):
+    """Convert Labelme to COCO dataset format"""
+
+    def __init__(self, anno_dir, image_dir, class_name=[], init_id=None, use_kpt=False, kpt_names=[],
+                 kpt_shape=[]):
+        """
+        :param anno_dir:
+        :param image_dir:
+        :param class_name:
+        :param init_id:
+        :param use_kpt:
+        :param kpt_names: 关键点名称列表，
+                          当kpt_names.ndim=1时，要求所有类别且所有目标的关键点名称必须一致
+                          当kpt_names.ndim=2时，仅要求相同类别目标的关键点名称必须一致，且len(kpt_names)==len(self.class_name)
+        :param kpt_shape: 关键点的维度，默认是(17, 3)，对于coco-person,有17个关键点，3表示(x,y,v)）
+        """
+        super(Labelme2COCO, self).__init__(init_id=init_id)
+        print(anno_dir)
+        print(image_dir)
+        self.anno_dir = anno_dir
+        self.image_dir = image_dir
+        self.kpt_names = kpt_names
+        self.kpt_shape = kpt_shape
+        self.labelme = parser_labelme.LabelMeDataset(anno_dir=anno_dir,
+                                                     image_dir=image_dir,
+                                                     class_name=class_name,
+                                                     check_kpt=True,
+                                                     use_kpt=use_kpt,
+                                                     kpt_names=kpt_names,
+                                                     kpt_shape=kpt_shape,
+                                                     use_rgb=False,
+                                                     shuffle=False,
+                                                     check=False
+                                                     )
+
+    def build_instances_dataset(self, save_file, ):
+        """
+        构建COCO的目标检测和实例分割数据集
+        :param save_file: 输出COCO格式的文件
+        :return:
+        """
+        for index in tqdm(range(len(self.labelme))):
+            data_info = self.labelme.__getitem__(index)
+            contours = [[c] for c in data_info["points"]]
+            info = {"boxes": data_info["boxes"], "labels": data_info["names"],
+                    "contours": contours}
+            width, height = data_info["size"]
+            filename = data_info["image_file"]
+            self.addObjects(filename, info, width, height, num_joints=0)
+        build_coco.COCOTools.check_coco(self.coco)
+        self.save_coco(save_file)
+
+    def build_keypoints_dataset(self, save_file, skeleton=[]):
+        """
+        构建COCO的目标检测和实例分割数据集
+        :param save_file: 输出COCO格式的文件
+        :return:
+        """
+        for index in tqdm(range(len(self.labelme))):
+            data_info = self.labelme.__getitem__(index)
+            contours = [[c] for c in data_info["points"]]
+            keypoints = [c.reshape(-1) for c in data_info["keypoints"]]
+            info = {"boxes": data_info["boxes"], "labels": data_info["names"],
+                    "contours": contours, "keypoints": keypoints}
+            width, height = data_info["size"]
+            filename = data_info["image_file"]
+            self.addObjects(filename, info, width, height)
+        # 设置关键点的名称和skeleton
+        self.set_keypoints_category(kps_name=self.kpt_names[0], skeleton=skeleton, cat_id=0)
+        build_coco.COCOTools.check_coco(self.coco)
+        self.save_coco(save_file)
+
+    def save_coco(self, json_file):
+        """保存COCO数据集"""
+        super(Labelme2COCO, self).save_coco(json_file)
+
+
+def demo():
+    class_name = ["person"]
+    kpt_names = [["p0", "p1", "p2", "p3", "p4"]]
+    skeleton = [[0, 2], [2, 1], [2, 3], [3, 4]]
+    kpt_shape = [5, 3]
+    anno_dir = "/media/PKing/dev2/SDK/base-utils/data/labelme/images"
+    coco_ins_file = os.path.join(os.path.dirname(anno_dir), "coco_ins.json")
+    coco_kps_file = os.path.join(os.path.dirname(anno_dir), "coco_kps.json")
+    build = Labelme2COCO(image_dir=anno_dir,
+                         anno_dir=anno_dir,
+                         class_name=class_name,
+                         use_kpt=True,
+                         kpt_names=kpt_names,
+                         kpt_shape=kpt_shape
+                         )
+    # TODO 构建关键点检测数据集
+    build.build_keypoints_dataset(coco_kps_file, skeleton=skeleton)
+    # TODO 构建实例分割数据集
+    # build.build_instances_dataset(coco_ins_file)
+
+
+if __name__ == '__main__':
+    demo()

{pybaseutils-2.4.35 → pybaseutils-2.5.2}/pybaseutils/dataloader/parser_labelme.py

@@ -53,6 +53,10 @@ class LabelMeDataset(Dataset):
         :param check: 当class_name=None且check=True,将自动获取所有class
         :param check_kpt: True会检查关键点的完整性,图像中仅当所有目标都标注和对应的关键点(无漏标注),才返回数据
                            False不会检查关键点的完整性,,图像中只要标注的目标和对应的关键点(漏标注),会返回数据
+        :param kpt_names: 关键点名称列表，
+                          当kpt_names.ndim=1时，要求所有类别且所有目标的关键点名称必须一致
+                          当kpt_names.ndim=2时，仅要求相同类别目标的关键点名称必须一致，且len(kpt_names)==len(self.class_name)
+        :param kpt_shape: 关键点的维度，默认是(17, 3)，对于coco-person,有17个关键点，3表示(x,y,v)
         :param min_points: 当标注的轮廓点的个数小于min_points，会被剔除；负数不剔除
         :param kwargs: read_image: 是否读取图片，否则image=None
         """
@@ -537,6 +541,34 @@ def LabelMeDatasets(filename=None,
     return datasets
 
 
+def get_class_counts(datasets, save=""):
+    """
+    统计每个类别的样本数量
+    :param datasets: List[str] or ConcatDataset
+    :param save: 保存路径
+    :return: 每个类别的样本数量
+    """
+    from pybaseutils import dict_utils
+    if isinstance(datasets, list):
+        datasets = LabelMeDatasets(filename=None,
+                                  data_root=None,
+                                  anno_dir=datasets,
+                                  image_dir=None,
+                                  class_name=[],
+                                  check=False,
+                                  phase='val',
+                                  shuffle=False)
+    counts = {}
+    for i in tqdm(range(len(datasets))):
+        info = datasets.__getitem__(i)
+        for n in info["names"]:
+            counts[n] = counts.get(n, 0) + 1
+    counts = dict_utils.dict_sort(counts,use_key=False)
+    if save:
+        json_utils.save_json(os.path.join(save, "counts.json"), counts)
+    return counts
+
+
 def parser_labelme(anno_file, class_dict={}, size=()):
     """
     :param annotation:  labelme标注的数据
@@ -591,9 +623,9 @@ def example_for_keypoints():
                               image_dir=None,
                               class_name=class_name,
                               use_kpt=True,
+                              check_kpt=True,
                               kpt_names=kpt_names,
                               kpt_shape=kpt_shape,
-                              check_kpt=False,
                               check=False,
                               phase="val",
                               shuffle=False)
@@ -633,4 +665,10 @@ def example_for_segment():
 
 if __name__ == "__main__":
     # example_for_segment()
-    example_for_keypoints()
+    # example_for_keypoints()
+    datasets = [
+        '/home/PKing/edudata/dataset/AICQ/dataset/data-2026-10-06/images',
+        '/home/PKing/edudata/dataset/AICQ/dataset/data-2026-09-30/images',
+    ]
+    counts = get_class_counts(datasets)
+    print(counts)

{pybaseutils-2.4.35 → pybaseutils-2.5.2}/pybaseutils/dataloader/parser_voc.py

@@ -119,7 +119,7 @@ class VOCDataset(Dataset):
             class_set = labels.reshape(-1).tolist() + class_set
             class_set = list(set(class_set))
             if ignore_empty and (len(boxes) == 0 or len(labels) == 0):
-                print("empty annotation:{}".format(annotation_file))
+                # print("empty annotation:{}".format(annotation_file))
                 continue
             dst_ids.append(image_id)
         class_set = sorted(class_set)

{pybaseutils-2.4.35 → pybaseutils-2.5.2}/pybaseutils/dict_uils.py

@@ -7,4 +7,5 @@
 # @Brief  :
 # --------------------------------------------------------
 """
+# TODO 兼容之前dict_uils.py的异常
 from pybaseutils.dict_utils import *

{pybaseutils-2.4.35 → pybaseutils-2.5.2}/pybaseutils/dict_utils.py

@@ -14,8 +14,9 @@ from collections import defaultdict
 
 
 class Objects(object):
-    def __init__(self, boxes, label, score, names=[], size=[], segms=None, masks=None, keypt=None,
-                 traid=None, trace=None, color=[], image=None, file=""):
+    def __init__(self, boxes, label, score, names=[], size=[],
+                 segms=[], masks=None, keypt=[], value=[],
+                 traid=[], trace: defaultdict = None, color=[], image=None, file=""):
         self.boxes = boxes  # 目标框(xmin,ymin,xmax,ymax)
         self.score = score  # 目标框置信度
         self.label = label  # 目标框类别
@@ -25,12 +26,14 @@ class Objects(object):
         self.masks = masks  # 目标轮廓掩码
         self.segms = segms  # 目标轮廓[np.shape(n,2)]
         self.traid = traid  # 轨迹ID
-        self.trace = trace  # 轨迹的历史boxes
+        self.trace = trace  # 轨迹的历史数据
+        self.value = value  # 一些数值
+        self.depth = None  # 深度图
+        self.dists = []  # 深度值(segms/boxes区域)
         self.color = [(255, 0, 0)] * len(boxes)  # 目标框颜色
+        self.metas = defaultdict()  # 其他信息
         self.size = size  # 原始图像大小
         self.file = file  # 文件路径
-        self.depth = None  # 深度图
-        self.metas = defaultdict()  # 其他信息
 
     def update(self, **kwargs):
         """更新属性值"""

{pybaseutils-2.4.35 → pybaseutils-2.5.2}/pybaseutils/font_style/__init__.py

@@ -19,6 +19,7 @@ FONT_ROOT = [
     os.path.expanduser("~/.fonts"),  # 用户字体目录
     BASE_FONT,  # 用户字体目录
     "/usr/share/fonts",
+    "./data/fonts",
     "/usr/share/fonts/truetype",
     "/usr/share/fonts/opentype",
 ]

{pybaseutils-2.4.35 → pybaseutils-2.5.2}/pybaseutils/font_utils.py

@@ -243,6 +243,30 @@ def re_example():
     print(out)
 
 
+def align_display(s, align='C', width=20, pchar=" "):
+    """
+    对齐显示字符串
+    print("[{}] 下载进度".format(align_display(item, 'C')))
+    :param s: 要显示的字符串
+    :param align: 对齐方式，'L' 左对齐，'C' 居中对齐，'R' 右对齐
+    :param width: 目标显示宽度
+    :param pchar: 对齐字符，默认空格
+    :return: 对齐后的字符串
+    """
+    w = sum(2 if ord(c) > 127 else 1 for c in s)  # 计算字符串的显示宽度（中文等宽字符算2，英文算1）
+    total_pad = max(width - w, 0)
+    if align == 'L':
+        L = 0
+        R = total_pad
+    elif align == 'R':
+        L = total_pad
+        R = 0
+    else:  # 默认居中对齐
+        L = total_pad // 2
+        R = total_pad - L
+    return pchar * L + s + pchar * R
+
+
 if __name__ == "__main__":
     draw_font_example()
     # get_system_fonts()

{pybaseutils-2.4.35 → pybaseutils-2.5.2}/pybaseutils/geometry_tools.py

@@ -7,6 +7,7 @@
 # @Brief  :
 # --------------------------------------------------------
 """
+import math
 import numpy as np
 from scipy.spatial.distance import cdist
 
@@ -221,6 +222,48 @@ def compute_vector_angle(a, b, minangle=True):
     return angle
 
 
+def image2math_coord(pts, height=None):
+    """
+    将图像坐标系（x 向右，y 向下）转换为数学坐标系（x 向右，y 向上）笛卡尔坐标系。
+    :param pts: 形如 (N, 2) 的数组/列表，图像坐标点
+    :param height: 可选，图像高度。给出时按 (x, height - y) 翻转，
+                   结果仍落在图像矩形范围内；不给出时按 (x, -y) 翻转。
+    :return: (N, 2) 的 numpy 数组，数学坐标点
+    """
+    pts = np.asarray(pts, dtype=float)
+    if pts.ndim != 2 or pts.shape[1] != 2:  raise ValueError("pts 必须是形如 (N, 2) 的数组")
+    out = pts.copy()
+    if height is None:
+        out[:, 1] = -out[:, 1]  # y 取反，原点不变
+    else:
+        out[:, 1] = height - out[:, 1]  # 以图像底边为 y=0
+    return out
+
+
+def get_vector_angle(v1, v2, clockwise=True):
+    """
+    坐标系约定: 标准的数学坐标系（x 轴向右，y 轴向上）
+    计算从向量 v1 旋转到向量 v2 的角度。
+    :param v1: 形如 (x, y) 的元组/列表，表示平面向量
+    :param v2: 形如 (x, y) 的元组/列表，表示平面向量
+    :param clockwise: True 顺时针角度，False 逆时针角度
+    :return: 角度（度），范围 [0, 360)
+             注意: 0 和 360 是同一个方向，函数统一返回 0
+    """
+    x1, y1 = v1
+    x2, y2 = v2
+    # 零向量无法定义方向
+    if (x1 == 0 and y1 == 0) or (x2 == 0 and y2 == 0): return 0
+    dot = x1 * x2 + y1 * y2  # 点积
+    cross = x1 * y2 - y1 * x2  # 叉积（z 分量）
+    # 叉积 > 0 表示 v2 在 v1 的逆时针方向
+    # 顺时针：用 -cross；逆时针：用 +cross
+    signed_cross = -cross if clockwise else cross
+    angle = math.degrees(math.atan2(signed_cross, dot))
+    # 把 (-180, 180] 映射到 [0, 360)
+    return angle % 360
+
+
 def line_test():
     '''
     angle: 56.789092174788685