pybaseutils 2.4.3__tar.gz → 2.4.5__tar.gz

{pybaseutils-2.4.3 → pybaseutils-2.4.5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pybaseutils
-Version: 2.4.3
+Version: 2.4.5
 Summary: pybaseutils
 Home-page: https://github.com/PanJinquan/base-utils
 Author: PanJinquan

{pybaseutils-2.4.3 → pybaseutils-2.4.5}/pybaseutils/__init__.py

@@ -8,4 +8,4 @@
 # --------------------------------------------------------
 """
 
-__version__ = '2.4.3'
+__version__ = '2.4.5'

{pybaseutils-2.4.3 → pybaseutils-2.4.5}/pybaseutils/base/list_queue.py

@@ -9,6 +9,7 @@
 """
 import math
 import queue
+import time
 
 
 class Queue():
@@ -46,7 +47,10 @@ class Queue():
         :param timeout: 超时时间
         :return: 弹出的队列数据
         """
-        out = self.queue.get(block=block, timeout=timeout)
+        try:
+            out = self.queue.get(block=block, timeout=timeout)
+        except:
+            out = None
         return out
 
     def pop_items(self, nums=1, block=True, timeout=None):
@@ -57,7 +61,10 @@ class Queue():
         :param timeout: 超时时间
         :return: 弹出的队列数据
         """
-        items = [self.pop(block=block, timeout=timeout) for i in range(nums)]
+        items = []
+        for i in range(nums):
+            item = self.pop(block=block, timeout=timeout)
+            if item: items.append(item)
         return items
 
     def get(self, index=0):
@@ -93,12 +100,18 @@ class Queue():
         获取队列的窗口数据
         :param winsize: 窗口大小
         :param overlap: 窗口重叠率
+        :param block: 是否阻塞等待
+        :param timeout: 超时时间
         :return: 队列窗口数据
         """
-        size = int(winsize * overlap)
-        data1 = self.pop_items(nums=winsize - size, block=block, timeout=timeout)
-        data2 = self.get_items(nums=size)
-        return data1 + data2
+        while self.qsize() < winsize and block:  # 等待队列数据足够
+            time.sleep(0.05)
+        if self.qsize() >= winsize:
+            size = int(winsize * overlap)
+            data1 = self.pop_items(nums=winsize - size, block=block, timeout=timeout)
+            data2 = self.get_items(nums=size)
+            return data1 + data2
+        return []
 
 
 if __name__ == '__main__':

{pybaseutils-2.4.3 → pybaseutils-2.4.5}/pybaseutils/base64_utils.py

@@ -9,6 +9,8 @@
 """
 import sys
 import os
+
+import PIL.Image as Image
 import cv2
 import base64
 import numpy as np
@@ -64,8 +66,8 @@ def image2base64(image: np.ndarray, prefix=IMG_PREFIX, use_rgb=False) -> str:
         img = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)
     ext = prefix.split("/")
     # ext = "." + ext[1] if len(ext) == 2 else ".png" # TODO libpng error: bad parameters to zlib
-    ext = "." + ext[1] if len(ext) == 2 else ".jpg"
-    img = cv2.imencode(ext, img)[1]
+    # ext = "." + ext[1] if len(ext) == 2 else ".jpg"
+    img = cv2.imencode('.jpg', img)[1]
     bs64 = prefix + base64.b64encode(img).decode()
     return bs64
 
@@ -77,7 +79,7 @@ def file2base64(file, prefix=IMG_PREFIX) -> str:
     :param prefix: base64字符串前缀,用于表识字符串的类型
     :return:base64字符串
     """
-    bs64 = prefix + base64.b64encode(open(file, 'rb').read()).decode()
+    bs64 = prefix + base64.b64encode(open(file, 'rb').read()).decode("utf-8")
     return bs64
 
 
@@ -105,6 +107,8 @@ def array2base64(data: Any, prefix=IMG_PREFIX, use_rgb=False) -> Any:
     """
     if isinstance(data, np.ndarray) and data.dtype == np.uint8:
         return image2base64(data, prefix=prefix, use_rgb=use_rgb)
+    elif isinstance(data, Image.Image):
+        return image2base64(np.asarray(data), prefix=prefix, use_rgb=use_rgb)
     elif isinstance(data, np.ndarray):
         return data.tolist()
     elif isinstance(data, np.integer):
@@ -122,22 +126,25 @@ def array2base64(data: Any, prefix=IMG_PREFIX, use_rgb=False) -> Any:
     return data
 
 
-def base642array(data: Any, prefix=IMG_PREFIX, use_rgb=False) -> Any:
+def base642array(data: Any, prefix=IMG_PREFIX, use_rgb=False, img_type="numpy") -> Any:
     """
     反序列化:将输入数据含有base64字符串都解码为图像数据(ndarray)
     :param data: 输入数据
     :param prefix: base64字符串前缀,用于表识字符串的类型
     :param use_rgb: True:输入image是RGB的图像, False:输入image是BGR格式的图像
+    :param img_type: numpy:返回numpy图像, pil:返回PIL图像
     :return:
     """
     if isinstance(data, str) and prefix == data[0:len(prefix)]:
-        return base642image(data, prefix=prefix, use_rgb=use_rgb)
+        data = base642image(data, prefix=prefix, use_rgb=use_rgb)  # numpy image
+        if img_type == "pil": data = Image.fromarray(data)
+        return data
     elif isinstance(data, list):
         for i in range(len(data)):
-            data[i] = base642array(data[i], prefix=prefix, use_rgb=use_rgb)
+            data[i] = base642array(data[i], prefix=prefix, use_rgb=use_rgb, img_type=img_type)
     elif isinstance(data, dict):
         for k, v in data.items():
-            data[k] = base642array(v, prefix=prefix, use_rgb=use_rgb)
+            data[k] = base642array(v, prefix=prefix, use_rgb=use_rgb, img_type=img_type)
     return data
 
 
@@ -147,13 +154,14 @@ if __name__ == "__main__":
     from pybaseutils import file_utils, image_utils
 
     image_dir = "/home/PKing/Downloads/image"
-    image_dir = "/media/PKing/dev1/SDK/base-utils/data/labelme/JPEGImages"
     image_list = file_utils.get_images_list(image_dir)
     for image_file in image_list:
         print(image_file)
         src = image_utils.read_image(image_file, use_rgb=True)
-        image_utils.show_image("src", src, delay=10)
-        data = {"image": src, "file": image_file}
+        image_base64 = image2base64(src, prefix="data:image/jpeg;base64,", )
+        img = Image.fromarray(src)
+        data = {"image": img, "file": image_file}
         data = serialization(data)
         data = deserialization(data)
-        print(data.keys())
+        dst = data["image"]
+        image_utils.show_image("src", src, delay=0)

{pybaseutils-2.4.3 → pybaseutils-2.4.5}/pybaseutils/converter/build_labelme.py

@@ -12,7 +12,7 @@ from tqdm import tqdm
 from pybaseutils import image_utils, file_utils, json_utils
 
 
-def save_labelme(out_root, image_file, points, names, class_dict, group=None, image=None, prefix="",
+def save_labelme(out_root, image_file, points, names, class_dict={}, group=None, image=None, prefix="",
                  index=0, vis=False, delay=0):
     """
     :param out_root:   输出根目录
@@ -39,15 +39,19 @@ def save_labelme(out_root, image_file, points, names, class_dict, group=None, im
     image_id, postfix = file_utils.split_postfix(image_name)
     if prefix:
         flag_ = file_utils.get_time(format="p") if index < 0 else f"{index:0=5d}"
-        image_name = f"{prefix}_{flag_}.{postfix}"
+        image_name = f"{prefix}_{flag_}.jpg"
     image_id, postfix = file_utils.split_postfix(image_name)
     json_file = file_utils.create_dir(out_root, "images", f"{image_id}.json")
     file_path = file_utils.create_dir(out_root, "images", f"{image_name}")
     maker_labelme(json_file, points, names, image_name, group=group, image_size=(w, h), image_bs64=None)
-    file_utils.copy_file(image_file, file_path)
+    if isinstance(image, np.ndarray):
+        cv2.imwrite(file_path, image)
+    else:
+        file_utils.copy_file(image_file, file_path)
     if len(points) == 0:
         print("points is empty,file={}".format(file_path))
 
+
 def maker_labelme(json_file, points, labels, image_name, image_size, group=None, image_bs64=None, keypoints=[]):
     """
     制作label数据格式

{pybaseutils-2.4.3 → pybaseutils-2.4.5}/pybaseutils/converter/convert_yolo2labelme.py

@@ -73,6 +73,7 @@ def convert_yolo2labelme(data_root,
                          class_name=None,
                          class_dict={},
                          task="",
+                         check=False,
                          prefix="",
                          max_num=-1,
                          vis=True,
@@ -93,7 +94,7 @@ def convert_yolo2labelme(data_root,
                                       image_dir=None,
                                       class_name=class_name,
                                       task=task,
-                                      check=False,
+                                      check=check,
                                       phase="val",
                                       shuffle=False)
     print("have num:{}".format(len(dataset)))

{pybaseutils-2.4.3 → pybaseutils-2.4.5}/pybaseutils/coords_utils.py

@@ -198,8 +198,8 @@ def shrink_polygon_pyclipper(polygon, ratio):
     使用Polygon库计算多边形区域的周长和面积，使用pyclipper库进行shrink
     https://blog.csdn.net/shyjhyp11/article/details/126396170
     https://www.cnblogs.com/01black-white/p/15292193.html
-    :param polygon: 
-    :param ratio: 
+    :param polygon:  (n,2)
+    :param ratio: 缩放的比例，如果大于1是放大,小于1是缩小
     :return: 
     """
     from shapely.geometry import Polygon
@@ -219,6 +219,9 @@ def shrink_polygon_pyclipper(polygon, ratio):
     return shrinked
 
 
+extend_polygon = shrink_polygon_pyclipper
+
+
 def get_square_boxes(boxes, use_max=True, use_mean=False, baseline=-1):
     """
     将boxes转换为正方形的boxes
@@ -499,7 +502,8 @@ def demo_for_augment():
     for i in range(1000):
         boxes = extend_xyxy(boxes, scale=[1.2, 1.2])
         dst_image, dst_label, dst_boxes = augment(image, boxes.copy(), labels.copy())
-        image = show_targets_image("detd", image, dst_label, dst_boxes, center2bboxes=True, untranspose=False, waitKey=0)
+        image = show_targets_image("detd", image, dst_label, dst_boxes, center2bboxes=True, untranspose=False,
+                                   waitKey=0)
 
 
 if __name__ == "__main__":

{pybaseutils-2.4.3 → pybaseutils-2.4.5}/pybaseutils/cvutils/video_utils.py

@@ -1,13 +1,16 @@
 # -*- coding: utf-8 -*-
 """
-    @Author : Pan
-    @E-mail : 390737991@qq.com
-    @Date   : 2022-05-24 16:46:51
-    @Brief  :
+# --------------------------------------------------------
+# @Author : Pan
+# @E-mail :
+# @Date   : 2025-12-12 14:23:50
+# @Brief  :
+# --------------------------------------------------------
 """
 import os
 import cv2
 import numpy as np
+import math
 from typing import Callable
 from tqdm import tqdm
 from pybaseutils import image_utils, file_utils
@@ -40,7 +43,7 @@ def video2gif(video_file, gif_file=None, func=None, interval=1, fps=-1, use_pil=
     name = os.path.basename(video_file).split(".")[0]
     if not gif_file:  gif_file = os.path.join(os.path.dirname(video_file), name + ".gif")
     if not os.path.exists(gif_file): file_utils.create_file_path(gif_file)
-    width, height, num_frames, _fps = get_video_info(video_file)
+    width, height, num_frames, _fps = get_video_info(video_file, **kwargs)
     video_cap = video_iterator(video_file, save_video=None, interval=interval, **kwargs)
     frames = []
     for data_info in video_cap:
@@ -56,29 +59,98 @@ def video2gif(video_file, gif_file=None, func=None, interval=1, fps=-1, use_pil=
         image_utils.frames2gif_by_imageio(frames, gif_file=gif_file, fps=fps, loop=0)
 
 
-def video2frames(video_file, out_dir=None, func=None, interval=1, start=0, end=-1, vis=True, delay=10):
+def get_video_sampling(freq, time, fps, random=False):
     """
+    获得抽帧时间点和索引
+    :param freq: 抽帧频率
+    :param time: time: 开始播放时间time[0]，结束播放时间time[1]，单位秒S
+    :param fps: 视频帧率FPS
+    :param random: 是否随机抽帧(在每一秒内随机抽freq帧)
+    :return:
+    """
+    if freq <= 0: freq = fps
+    if random:
+        times = []
+        tange = np.arange(time[0], math.ceil(time[1]) + 1e-6, 1)
+        for i in range(len(tange) - 1):
+            t = np.random.uniform(tange[i], tange[i + 1], size=freq)
+            t = np.sort(t)
+            times.append(t)
+        times = np.concatenate(times)
+        # times = np.sort(np.random.uniform(time[0], time[1], size=freq))
+    else:
+        times = np.arange(time[0], time[1], 1 / freq)
+    index = times * fps
+    index = index.astype(int)
+    return times, index
+
+
+def load_video(video_file, time=(0, -1), freq=0, size=(), use_rgb=False, random=False, **kwargs):
+    """
+    加载视频文件，返回视频抽帧图像数据列表
+    :param video_file:
+    :param time: 设置时间范围，默认值为(0, -1)，表示播放所有视频帧
+    :param freq: 抽帧频率，默认值为视频帧率FPS
+    :param size: 帧图像大小，默认值为视频原始大小
+    :param use_rgb:
+    :param kwargs:
+    :return:
+    """
+    time = list(time)
+    video_cap = get_video_capture(video_file)
+    width, height, numFrames, fps = get_video_info(video_cap, **kwargs)
+    if time[1] < 0: time[1] = numFrames / fps
+    video_time, video_index = get_video_sampling(freq, time, fps, random=random)
+    frames = []
+    for t, count in zip(video_time, video_index):
+        video_cap.set(cv2.CAP_PROP_POS_FRAMES, count)
+        ret, frame = video_cap.read()
+        if not ret: break
+        if size: frame = image_utils.resize_image(frame, size=size)
+        if use_rgb: frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
+        h, w = frame.shape[:2]
+        data_info = {"count": count, "time": t, "frame": frame, "w": w, "h": h, "fps": fps}
+        frames.append(data_info)
+    video_cap.release()
+    return frames
+
+
+def video2frames(video_file, out_dir=None, task: Callable = None, interval=1, size=(), freq=0,
+                 vis=False, delay=10, **kwargs):
+    """
+    video_iterator(video_file: int | str, save_video: str or int = None, interval=1, size=(), freq=0,
+                   task: Callable = None, vis=False, **kwargs):
     视频抽帧图像
     :param video_file: 视频文件
-    :param out_dir: 保存抽帧图像的目录
-    :param func: 回调函数，对每一帧图像进行处理
+    :param out_dir: 保存抽帧图像的目录,os.path.join(os.path.dirname(video_file), name)
+                    当out_dir为None时，返回抽帧图像数据列表
+    :param task: 回调函数，对每一帧图像进行处理
     :param interval: 保存间隔
     :param vis: 是否可视化显示
     :return:
     """
-    name = os.path.basename(video_file).split(".")[0]
-    if not out_dir:  out_dir = os.path.join(os.path.dirname(video_file), name)
-    video_cap = video_iterator(video_file, save_video=None, interval=interval, start=start, end=end)
+    video_cap = video_iterator(video_file, save_video=None, interval=interval, size=size, freq=freq, **kwargs)
+    filename = None
+    if out_dir:
+        name = os.path.basename(video_file).split(".")[0]
+        # out_dir = os.path.join(os.path.dirname(video_file), name)
+        prefix = kwargs.get("prefix", "")
+        filename = "{}_{}".format(prefix, name) if prefix else name
+    frames = []
     for data_info in video_cap:
         frame = data_info["frame"]
         count = data_info["count"]
-        if func: frame = func(frame)
+        if task: frame = task(frame)
         if vis: image_utils.cv_show_image("frame", frame, use_rgb=False, delay=delay)
-        frame_file = os.path.join(out_dir, "{}_{:0=4d}.jpg".format(name, count))
-        cv2.imwrite(frame_file, frame)
+        if filename:
+            frame_file = os.path.join(out_dir, "{}_{:0=5d}.jpg".format(filename, count))
+            cv2.imwrite(frame_file, frame)
+            data_info["file"] = frame_file
+        frames.append(data_info)
+    return frames
 
 
-def video2frames_similarity(video_file, out_dir=None, func=None, interval=1, thresh=0.3, vis=True):
+def video2frames_similarity(video_file, out_dir=None, func=None, interval=1, thresh=0.3, vis=True, **kwargs):
     """
     视频抽帧图像
     :param video_file: 视频文件
@@ -92,10 +164,11 @@ def video2frames_similarity(video_file, out_dir=None, func=None, interval=1, thr
     name = os.path.basename(video_file).split(".")[0]
     if not out_dir:  out_dir = os.path.join(os.path.dirname(video_file), name)
     video_cap = get_video_capture(video_file)
-    width, height, num_frames, fps = get_video_info(video_cap)
+    width, height, num_frames, fps = get_video_info(video_cap, **kwargs)
     if not os.path.exists(out_dir): os.makedirs(out_dir)
     count = 0
     last_frame = None
+    frame_files = []
     while True:
         if count % interval == 0:
             # 设置抽帧的位置
@@ -110,6 +183,7 @@ def video2frames_similarity(video_file, out_dir=None, func=None, interval=1, thr
                 frame_file = os.path.join(out_dir, "{}_{:0=4d}.jpg".format(name, count))
                 last_frame = curr_frame.copy()
                 cv2.imwrite(frame_file, curr_frame)
+                frame_files.append(frame_file)
             if vis:
                 text = "TH={},diff={:3.3f}".format(thresh, diff)
                 image = image_utils.draw_text(curr_frame, point=(10, 70), color=(0, 255, 0),
@@ -118,6 +192,7 @@ def video2frames_similarity(video_file, out_dir=None, func=None, interval=1, thr
         count += 1
     video_cap.release()
     cv2.destroyAllWindows()
+    return frame_files
 
 
 def frames2video(image_dir, video_file=None, func=None, size=None, postfix=["*.png", "*.jpg"], interval=1, fps=30,
@@ -160,7 +235,7 @@ def frames2video(image_dir, video_file=None, func=None, size=None, postfix=["*.p
     cv2.destroyAllWindows()
 
 
-def video2video(video_file: int or str, save_video: str or int, interval=1, task: Callable = None,
+def video2video(video_file: int or str, save_video: str or int, interval=1, size=(), freq=0, task: Callable = None,
                 vis=True, **kwargs):
     """
     转换视频格式
@@ -169,13 +244,14 @@ def video2video(video_file: int or str, save_video: str or int, interval=1, task
     :param interval: 间隔
     :return:
     """
-    video_capture(video_file=video_file, save_video=save_video, interval=interval, task=task, vis=vis, **kwargs)
+    video_capture(video_file=video_file, save_video=save_video, interval=interval, size=size, freq=freq, task=task,
+                  vis=vis, **kwargs)
 
 
 convert_video_format = video2video
 
 
-def resize_video(video_file, save_video, size=(), start=0, interval=1, vis=True, delay=20):
+def resize_video(video_file, save_video, size=(), start=0, interval=1, vis=True, delay=20, **kwargs):
     """
     转换视频格式
     :param video_file: *.avi,*.mp4,...
@@ -185,7 +261,7 @@ def resize_video(video_file, save_video, size=(), start=0, interval=1, vis=True,
     :return:
     """
     video_cap = get_video_capture(video_file)
-    width, height, num_frames, fps = get_video_info(video_cap)
+    width, height, num_frames, fps = get_video_info(video_cap, **kwargs)
     frame = np.zeros(shape=(height, width), dtype=np.uint8)
     frame = image_utils.resize_image(frame, size=size)
     height, width = frame.shape[:2]
@@ -207,8 +283,8 @@ def resize_video(video_file, save_video, size=(), start=0, interval=1, vis=True,
     video_writer.release()
 
 
-def video_capture(video_file: int or str, save_video: str or int = None, interval=1, freq=0, task: Callable = None,
-                  vis=True, **kwargs):
+def video_capture(video_file: int or str, save_video: str or int = None, interval=1, size=(), freq=0,
+                  task: Callable = None, vis=True, **kwargs):
     """
     读取摄像头或者视频流
     :param video_file: String 视频文件，如*.avi,*.mp4,...
@@ -223,16 +299,22 @@ def video_capture(video_file: int or str, save_video: str or int = None, interva
     :return:
     """
     video_file = file_utils.str2number(video_file)
-    if isinstance(video_file, str): assert os.path.exists(video_file), f"video_file={video_file}"
-    video_cap = image_utils.get_video_capture(video_file)
-    w, h, num_frames, fps = image_utils.get_video_info(video_cap)
-    start = int(kwargs.get("start", 0) * fps)
-    end = int(kwargs.get("end", num_frames / fps) * fps) if fps > 0 else 0
-    if num_frames <= 0: num_frames = end
-    end = min(end, num_frames)  # TODO 当num_frames<0时，使用0<end<count继续播放
+    if isinstance(video_file, str) and os.path.isfile(video_file):
+        assert os.path.exists(video_file), f"video_file={video_file}"
+    video_cap = image_utils.get_video_capture(video_file, fps=None)
+    w, h, num_frames, fps = image_utils.get_video_info(video_cap, **kwargs)
+    time = kwargs.get("time", tuple())  # TODO 开始播放时间time[0]，结束播放时间time[1]，单位秒S
+    clip = kwargs.get("clip", tuple())  # TODO 开始播放位置index[0]，结束播放位置index[1]
+    start, end = 0, -1
+    if time and fps > 0:
+        start, end = int(time[0] * fps), int(time[1] * fps)
+    elif clip:
+        start, end = int(clip[0]), int(clip[1])
+    end = min(end, num_frames) if end > 0 else num_frames  # TODO 当num_frames<0时，使用0<end<count继续播放
     interval = fps if interval == -1 else interval  # 当interval=-1，表示interval=fps,即一秒一帧
     interval = int(fps / freq) if freq > 0 else interval
     save_fps = max(kwargs.get("speed", 1) * fps // interval, 1)
+    save_fps = kwargs.get("save_fps", save_fps)
     count = 0
     video_writer = None
     while True:
@@ -242,6 +324,7 @@ def video_capture(video_file: int or str, save_video: str or int = None, interva
             # if isinstance(video_file, str): video_cap.set(cv2.CAP_PROP_POS_FRAMES, count)
             # ret, frame = video_cap.read()
             if not ret or 0 < end < count or frame is None: break
+            if size: frame = image_utils.resize_image(frame, size=size)
             if task: frame = task(frame, **kwargs)
             h, w = frame.shape[:2]
             if vis: image_utils.cv_show_image(kwargs.get("title", "video"), frame, delay=kwargs.get("delay", 10))
@@ -255,13 +338,13 @@ def video_capture(video_file: int or str, save_video: str or int = None, interva
         video_writer.release()
 
 
-def video_iterator(video_file: int or str, save_video: str or int = None, interval=1, size=(), freq=0,
+def video_iterator(video_file: int | str, save_video: str or int = None, interval=1, size=(), freq=0,
                    task: Callable = None, vis=False, **kwargs):
     """
     读取摄像头或者视频流迭代器
     Usage:
         from pybaseutils.cvutils import video_utils
-        video_cap = video_utils.video_iterator(video_file, save_video, start=4, end=10)
+        video_cap = video_utils.video_iterator(video_file, save_video, time=(4, 10))
         for data_info in video_cap:
             frame = data_info["frame"]
             ...
@@ -275,43 +358,61 @@ def video_iterator(video_file: int or str, save_video: str or int = None, interv
     :param kwargs:回调函数输入参数,
                  delay: 控制显示延时,默认10S
                  title: 控制显示窗口名，默认video
-                 start: 开始播放时间，单位S
-                 end: 结束播放时间，单位S
-                 speed: 播放速度
+                 time: 开始播放时间time[0]，结束播放时间time[1]，单位秒S
+                 clip: 开始播放位置index[0]，结束播放位置index[1]
+                 speed: 保存视频的播放速度，默认是播放帧率
+                 save_fps: 保存视频的帧率，默认是播放帧率
     :return: frame, count, w, h, fps =data_info['frame'],data_info['count'],data_info['w'],data_info['h'],data_info['fps']
+             当输入是视频文件时，返回视频偏移量time和duration都是播放时间，单位S，差异不大
+             当输入是摄像头时， 返回视频偏移量time是视频播放时间，duration是根据count计算的播放的时间
     """
     video_file = file_utils.str2number(video_file)
-    if isinstance(video_file, str): assert os.path.exists(video_file), f"video_file={video_file}"
+    if isinstance(video_file, str) and os.path.isfile(video_file):
+        assert os.path.exists(video_file), f"video_file={video_file}"
     video_cap = image_utils.get_video_capture(video_file, fps=None)
-    w, h, num_frames, fps = image_utils.get_video_info(video_cap)
-    start = int(kwargs.get("start", 0) * fps)
-    end = int(kwargs.get("end", num_frames / fps) * fps) if fps > 0 else 0
+    w, h, num_frames, fps = image_utils.get_video_info(video_cap, **kwargs)
+    time = kwargs.get("time", tuple())  # TODO 开始播放时间time[0]，结束播放时间time[1]，单位秒S
+    clip = kwargs.get("clip", tuple())  # TODO 开始播放位置index[0]，结束播放位置index[1]
+    start, end = 0, -1
+    if time and fps > 0:
+        start, end = int(time[0] * fps), int(time[1] * fps)
+    elif clip:
+        start, end = int(clip[0]), int(clip[1])
     end = min(end, num_frames) if end > 0 else num_frames  # TODO 当num_frames<0时，使用0<end<count继续播放
     interval = fps if interval == -1 else interval  # 当interval=-1，表示interval=fps,即一秒一帧
     interval = int(fps / freq) if freq > 0 else interval
     save_fps = max(kwargs.get("speed", 1) * fps // interval, 1)
+    save_fps = kwargs.get("save_fps", save_fps)
     count = 0
     video_writer = None
     use_fast = kwargs.get("use_fast", False)
     data_info = {}
+    t0 = -1
     while True:
         ret, frame = (False, None) if use_fast else video_cap.read()
+        t = video_cap.get(cv2.CAP_PROP_POS_MSEC) / 1000  # 获得视频偏移量毫秒为单位
+        if t0 < 0 < t: t0 = t
+        if t0 > 0: t = t - t0  # 获得视频偏移量毫秒为单位
+        finish = 0 < end <= (count + interval)
         if count % interval == 0 and count >= start:
             # TODO 设置抽帧的位置，但某些格式视频容易出现问题
             if use_fast and isinstance(video_file, str):
                 video_cap.set(cv2.CAP_PROP_POS_FRAMES, count)
                 ret, frame = video_cap.read()
-            if not ret or 0 < end < count or frame is None: break
+            if not ret or 0 < end <= count or frame is None: break
             if size: frame = image_utils.resize_image(frame, size=size)
             if task: frame = task(frame, **kwargs)
-            t = round(count / fps, 3)
-            data_info = {"count": count, "time": t, "frame": frame, "w": w, "h": h, "fps": fps, "offset": count}
+            h, w = frame.shape[:2]
+            d = round(count / fps, 3)  # TODO 通过fps计算播放时间
+            t = round(t, 3)  # 通过视频偏移量计算播放时间
+            data_info = {"count": count, "time": t, "frame": frame, "w": w, "h": h, "fps": fps,
+                         "finish": finish, 'duration': d}
             # TODO 返回data_info
             yield data_info
             frame = data_info["frame"]
-            h, w = frame.shape[:2]
             if vis: image_utils.cv_show_image(kwargs.get("title", "video"), frame, delay=kwargs.get("delay", 10))
             if save_video:
+                h, w = frame.shape[:2]
                 if not video_writer: video_writer = image_utils.get_video_writer(save_video, w, h, save_fps)
                 video_writer.write(frame)
         count += 1
@@ -340,10 +441,12 @@ def rotation_task(frame, **kwargs):
 
 
 if __name__ == "__main__":
-    video_file = "/home/dm/nasdata/release/CSDN/双目测距Demo视频(Python).MP4"
+    # video_file = "/home/dm/nasdata/release/CSDN/双目测距Demo视频(Python).MP4"
     # video_file = "/home/dm/视频/双目测距Demo视频(Python).mp4"
     # dst_file = "/home/dm/视频/双目测距Demo视频(Python)1.mp4"
     # video2frames(video_file, interval=10, vis=True)
     # frames2video(image_dir, interval=1, vis=True)
-    video2gif(video_file, interval=15, func=resize_task, fps=3, use_pil=False, vis=True)
+    # video2gif(video_file, interval=15, func=resize_task, fps=3, use_pil=False, vis=True)
     # video2video(video_file, dst_file, vis=True)
+    video_file = "/home/PKing/Videos/aije-data/检查绝缘棒.mp4"
+    load_video(video_file, time=(0, -1), freq=2, size=(), use_rgb=False)

{pybaseutils-2.4.3 → pybaseutils-2.4.5}/pybaseutils/dataloader/base_dataset.py

@@ -329,7 +329,7 @@ def cat_targets(objs: list, keys=[], out_info={}):
     return out_info
 
 
-def get_targets(obj_info: dict, targets=[], key='label', keys=[]):
+def get_targets(obj_info: dict, targets=[], key='names', keys=[]):
     """
     从obj_info查找符合条件的目标，支持正则表达式
     :param obj_info:
@@ -338,9 +338,9 @@ def get_targets(obj_info: dict, targets=[], key='label', keys=[]):
     :param keys: 用于指定返回的keys
     :return:
     """
-    if not obj_info: return {}
+    output = {k: [] for k in keys}
+    if not obj_info: return output
     label = obj_info[key]
-    output = {}
     for i in range(len(label)):
         matches = text_utils.find_match_texts(texts=[label[i]], pattern=targets, org=True)
         if len(matches) == 0: continue