pybaseutils 1.2.0__tar.gz → 2.0.0__tar.gz

{pybaseutils-1.2.0 → pybaseutils-2.0.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: pybaseutils
-Version: 1.2.0
+Version: 2.0.0
 Summary: pybaseutils
 Home-page: https://github.com/PanJinquan/base-utils
 Author: PanJinquan

{pybaseutils-1.2.0 → pybaseutils-2.0.0}/pybaseutils/__init__.py

@@ -5,4 +5,4 @@
     @Date   : 2019-05-07 17:40:27
 """
 
-__version__ = '1.2.0'
+__version__ = '2.0.0'

pybaseutils-2.0.0/pybaseutils/converter/build_cvat.py

@@ -0,0 +1,53 @@
+# -*-coding: utf-8 -*-
+"""
+    @Author : Pan
+    @E-mail : 390737991@qq.com
+    @Date   : 2022-09-05 19:48:52
+    @Brief  :
+"""
+import os
+import numpy as np
+import cv2
+from pybaseutils import image_utils, file_utils
+import xmltodict
+
+
+def maker_cvat(xml_file, points, labels, image_name, image_size):
+    """
+    制作CVAT数据格式
+    :param xml_file: 保存xml文件路径
+    :param points: (num_labels,num_points,2), points = image_utils.boxes2polygons(boxes)
+    :param labels: (num_labels,)
+    :param image_name: 图片名称，如果存在则进行拷贝到json_file同一级目录
+    :param image_size: (W,H)
+    :return:
+    """
+    assert len(points) == len(labels)
+    file_utils.create_file_path(xml_file)
+    objects = []
+    for idx, (point, label) in enumerate(zip(points, labels)):
+        if isinstance(point, np.ndarray): point = point.tolist()
+        if not isinstance(point[0], list): point = [point]
+        pt = [{'x': p[0], 'y': p[1]} for p in point]
+        item = {'name': label,
+                'deleted': '0',
+                'verified': '0',
+                'occluded': 'no',
+                'date': None,
+                'id': idx,
+                'parts': {'hasparts': None, 'ispartof': None},
+                'polygon': {'pt': pt, 'username': None},
+                'attributes': None}
+        objects.append(item)
+    data_info = {
+        "annotation": {
+            'filename': os.path.basename(image_name),
+            'folder': "",
+            'source': {'sourceImage': None, 'sourceAnnotation': 'Datumaro'},
+            'imagesize': {'nrows': image_size[1], 'ncols': image_size[0]},
+            'object': objects
+        }}
+    xml_info = xmltodict.unparse(data_info)
+    with open(xml_file, 'w') as xml_file:
+        xml_file.write(xml_info)
+    return data_info

{pybaseutils-1.2.0 → pybaseutils-2.0.0}/pybaseutils/converter/convert_cvat2labelme.py

@@ -41,13 +41,13 @@ def parser_annotations(xml_file):
 def convert_cvat2labelme(anno_dir, image_dir="", vis=False):
     """
     将CVAT标注格式(LabelMe 3.0)转换labelme通用格式
-    :param anno_dire:
-    :param image_dir:
+    :param anno_dir: 标注文件目录，输出json文件与anno_dir同目录
+    :param image_dir: 图片文件目录
     :param vis:
     :return:
     """
     xml_list = file_utils.get_files_lists(anno_dir, postfix=["*.xml"])
-    for xml_file in xml_list:
+    for xml_file in tqdm(xml_list):
         image_name, annos, w, h = parser_annotations(xml_file)
         points = [an['points'] for an in annos if len(an['points']) > 0]
         labels = [an['label'] for an in annos if an['label']]

pybaseutils-2.0.0/pybaseutils/converter/convert_labelme2cvat.py

@@ -0,0 +1,39 @@
+# -*-coding: utf-8 -*-
+"""
+    @Author : Pan
+    @E-mail : 390737991@qq.com
+    @Date   : 2022-10-12 16:21:55
+    @Brief  :
+"""
+import os
+import numpy as np
+import cv2
+from tqdm import tqdm
+from pybaseutils.dataloader import parser_voc, parser_labelme
+from pybaseutils.converter import build_voc, build_labelme, build_cvat
+from pybaseutils import file_utils, image_utils, json_utils
+
+
+def convert_labelme2cvat(anno_dir, image_dir="", vis=False):
+    """
+    将labelme文件转换为CVAT标注格式(LabelMe 3.0)
+    :param anno_dir: 标注文件目录，
+    :param image_dir: 图片文件目录，输出xml文件与anno_dir同目录
+    :param vis:
+    :return:
+    """
+    image_list = file_utils.get_images_list(image_dir)
+    for image_file in tqdm(image_list):
+        image_name = os.path.basename(image_file)
+        json_file = os.path.join(anno_dir, image_name.split(".")[0] + ".json")
+        xml_file = os.path.join(image_dir, image_name.split(".")[0] + ".xml")
+        boxes, labels, points, groups = parser_labelme.parser_labelme(json_file, class_dict={})
+        image = cv2.imread(image_file)
+        h, w = image.shape[:2]
+        build_cvat.maker_cvat(xml_file, points, labels, image_name, image_size=[w, h])
+
+
+if __name__ == "__main__":
+    image_dir = "/home/PKing/Downloads/dataset-label/image"
+    anno_dir = "/home/PKing/Downloads/dataset-label/json"
+    convert_labelme2cvat(image_dir=image_dir, anno_dir=anno_dir, vis=False)

{pybaseutils-1.2.0 → pybaseutils-2.0.0}/pybaseutils/converter/convert_labelme2voc.py

@@ -106,7 +106,7 @@ class Labelme2VOC(object):
         for i, item in enumerate(objects):
             label = item["name"]
             box = item["bndbox"]
-            img = image_utils.get_bbox_crop(image, bbox=box)
+            img = image_utils.get_box_crop(image, box=box)
             file = os.path.join(out_dir, str(label), "{}_{}_{:0=3d}.jpg".format(image_id, label, i))
             file_utils.create_file_path(file)
             cv2.imwrite(file, img)

{pybaseutils-1.2.0 → pybaseutils-2.0.0}/pybaseutils/coords_utils.py

@@ -328,7 +328,7 @@ def show_image(name, image, boxes, labels, center2bboxes=False, untranspose=Fals
         boxes = cxcywh2xyxy(boxes)
     if untranspose:
         image = image_utils.untranspose(image)
-    image = image_utils.show_image_bboxes_text(name, image, boxes, labels, delay=waitKey)
+    image = image_utils.show_image_boxes_texts(name, image, boxes, labels, delay=waitKey)
     return image
 
 

{pybaseutils-1.2.0 → pybaseutils-2.0.0}/pybaseutils/cvutils/video_utils.py

@@ -244,17 +244,8 @@ def resize_video(video_file, save_video, size=(), start=0, interval=1, vis=True,
     video_writer.release()
 
 
-def video_task(frame, **kwargs):
-    # TODO
-    delay = kwargs.get("delay", 0)
-    title = kwargs.get("title", "image")
-    cv2.imshow(title, frame)
-    cv2.moveWindow(title, 0, 0)
-    cv2.waitKey(delay)
-    return frame
-
-
-def video_capture(video_file: int or str, save_video: str = None, interval=1, task: Callable = video_task, **kwargs):
+def video_capture(video_file: int or str, save_video: str or int = None, interval=1, 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,...
@@ -267,66 +258,86 @@ def video_capture(video_file: int or str, save_video: str = None, interval=1, ta
                  title: 控制显示窗口名
     :return:
     """
-    # cv2.moveWindow("test", 1000, 100)
-    video_cap = get_video_capture(video_file)
-    width, height, num_frames, fps = get_video_info(video_cap)
+    video_cap = image_utils.get_video_capture(video_file)
+    width, height, num_frames, fps = image_utils.get_video_info(video_cap)
+    # video_writer = get_video_writer(save_video, width, height, fps)
     video_writer = None
-    # fps = max((fps + interval - 1) // interval, 2)
     fps = max(fps // interval, 2)
-    if save_video: video_writer = get_video_writer(save_video, width, height, fps)
-    # freq = int(fps / detect_freq)
-    count = 0
+    count = int(kwargs.get("start", 0) * fps)
     while True:
-        if count % interval == 0:
+        # if count % interval == 0:
+        if count % interval == 0 and count >= 0:
             # 设置抽帧的位置
             if isinstance(video_file, str): video_cap.set(cv2.CAP_PROP_POS_FRAMES, count)
             isSuccess, frame = video_cap.read()
             if not isSuccess or 0 < num_frames < count: break
             if task: frame = task(frame, **kwargs)
+            height, width = frame.shape[:2]
+            if vis: image_utils.cv_show_image(kwargs.get("title", "video"), frame, delay=kwargs.get("delay", 10))
             if save_video:
+                if not video_writer: video_writer = image_utils.get_video_writer(save_video, width, height, fps)
                 video_writer.write(frame)
         count += 1
     video_cap.release()
+    if video_writer:
+        print("save video:{}".format(save_video))
+        video_writer.release()
 
 
-class CVVideo():
-    def __init__(self):
-        pass
-
-    def video_capture(self, video_file, save_video=None, interval=1):
-        """
-        start capture video
-        :param video_file: *.avi,*.mp4,...
-        :param save_video: *.avi
-        :param interval:
-        :return:
-        """
-        # cv2.moveWindow("test", 1000, 100)
-        video_cap = get_video_capture(video_file)
-        width, height, num_frames, fps = get_video_info(video_cap)
-        video_writer = None
-        if save_video: video_writer = get_video_writer(save_video, width, height, fps)
-        # freq = int(fps / detect_freq)
-        count = 0
-        while True:
-            if count % interval == 0:
-                # 设置抽帧的位置
-                if isinstance(video_file, str): video_cap.set(cv2.CAP_PROP_POS_FRAMES, count)
-                isSuccess, frame = video_cap.read()
-                if not isSuccess or 0 < num_frames < count: break
-                frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
-                frame = self.task(frame)
-                if save_video:
-                    video_writer.write(frame)
-            count += 1
-        video_cap.release()
-
-    def task(self, frame):
-        # TODO
-        cv2.imshow("image", frame)
-        cv2.moveWindow("image", 0, 0)
-        cv2.waitKey(10)
-        return frame
+def video_iterator(video_file: int or str, save_video: str or int = None, interval=1, 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)
+        for data_info in video_cap:
+            frame = data_info["frame"]
+            ...
+    :param video_file: String 视频文件，如*.avi,*.mp4,...
+                       Int 摄像头ID，如0，1，2
+    :param save_video: 保存task视频处理后的结果
+    :param interval: 抽帧处理间隔，当interval=-1，表示当interval=fps,即一秒一帧
+    :param task: 回调函数： def task(frame, **kwargs)
+    :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
+    :return: frame, count, w, h, fps =data_info['frame'],data_info['count'],data_info['w'],data_info['h'],data_info['fps']
+    """
+    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)
+    end = min(end, num_frames)
+    interval = fps if interval == -1 else interval  # 当interval=-1，表示interval=fps,即一秒一帧
+    save_fps = max(kwargs.get("speed", 1) * fps // interval, 1)
+    count = start
+    video_writer = None
+    while True:
+        # if count % interval == 0:
+        if count % interval == 0 and count >= 0:
+            # 设置抽帧的位置
+            if isinstance(video_file, str): video_cap.set(cv2.CAP_PROP_POS_FRAMES, count)
+            isSuccess, frame = video_cap.read()
+            if not isSuccess or 0 < end < count or frame is None: break
+            if task: frame = task(frame, **kwargs)
+            data_info = {"frame": frame, "count": count, "w": w, "h": h, "fps": fps}
+            # 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:
+                if not video_writer: video_writer = image_utils.get_video_writer(save_video, w, h, save_fps)
+                video_writer.write(frame)
+        count += 1
+    video_cap.release()
+    if video_writer:
+        print("save video:{}".format(save_video))
+        video_writer.release()
 
 
 def resize_task(frame, **kwargs):
@@ -342,7 +353,7 @@ def rotation_task(frame, **kwargs):
     angle = angle + angle[::-1]
     h, w = frame.shape[:2]
     frame = image_utils.image_rotation(frame, angle=alpha * angle[count % len(angle)])
-    frame = image_utils.get_bbox_crop(frame, bbox=[0, 70, w, h - 70])
+    frame = image_utils.get_box_crop(frame, box=[0, 70, w, h - 70])
     return frame
 
 

{pybaseutils-1.2.0 → pybaseutils-2.0.0}/pybaseutils/dataloader/parser_labelme.py

@@ -155,6 +155,7 @@ class LabelMeDataset(Dataset):
             image_ids = [os.path.basename(f) for f in image_ids]
         # assert os.path.exists(image_dir), Exception("no directory:{}".format(image_dir))
         # assert os.path.exists(anno_dir), Exception("no directory:{}".format(anno_dir))
+        assert len(image_ids) > 0, f"image_dir={image_dir} is empty"
         return data_root, anno_dir, image_dir, image_ids
 
     def __getitem__(self, index):
@@ -380,12 +381,12 @@ def parser_labelme(anno_file, class_dict={}, shape=None):
     return bboxes, labels, points, groups
 
 
-def show_target_image(image, bboxes, labels, points):
+def show_target_image(image, bboxes, labels, points, color=()):
     # image = image_utils.draw_image_bboxes_text(image, bboxes, labels, color=(255, 0, 0),
     #                                            thickness=2, fontScale=1.2, drawType="chinese")
     # image = image_utils.draw_landmark(image, points, color=(0, 255, 0))
     # image = image_utils.draw_key_point_in_image(image, points)
-    image = image_utils.draw_image_contours(image, points, labels)
+    image = image_utils.draw_image_contours(image, points, labels, color=color, thickness=1)
     image_utils.cv_show_image("det", image)
 
 

{pybaseutils-1.2.0 → pybaseutils-2.0.0}/pybaseutils/dataloader/parser_voc.py

@@ -147,6 +147,7 @@ class VOCDataset(Dataset):
 
         files = self.get_file_list(image_dir, postfix=file_utils.IMG_POSTFIX, basename=False)
         self.postfix = os.path.basename(files[0]).split(".")[-1] if files else "jpg"
+        assert len(image_ids) > 0, f"image_dir={image_dir} is empty"
         return data_root, anno_dir, image_dir, image_ids
 
     def convert_target(self, boxes, labels):

{pybaseutils-1.2.0 → pybaseutils-2.0.0}/pybaseutils/dataloader/parser_yolo.py

@@ -204,6 +204,7 @@ class YOLODataset(Dataset):
             image_ids = [os.path.basename(f) for f in image_ids]
         # assert os.path.exists(image_dir), Exception("no directory:{}".format(image_dir))
         # assert os.path.exists(anno_dir), Exception("no directory:{}".format(anno_dir))
+        assert len(image_ids) > 0, f"image_dir={image_dir} is empty"
         return data_root, anno_dir, image_dir, image_ids
 
     def __getitem__(self, index):

{pybaseutils-1.2.0 → pybaseutils-2.0.0}/pybaseutils/font_utils.py

@@ -122,7 +122,7 @@ def draw_font(text, style="楷体", scale=1.0, size=20, c1=(255, 255, 255), c2=(
         if np.sum(image) < 1: return None
         mask = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
         box = image_utils.get_mask_boundrect_cv(mask, binarize=True, shift=10)
-        image = image_utils.get_bbox_crop(image, box)
+        image = image_utils.get_box_crop(image, box)
         image = image_utils.resize_image_padding(image, size=(size, size), color=(0, 0, 0))
     else:
         image = image_utils.create_image(shape=(size, size, 3), color=c2)