pybaseutils 2.4.11__tar.gz → 2.4.15__tar.gz

{pybaseutils-2.4.11 → pybaseutils-2.4.15}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pybaseutils
-Version: 2.4.11
+Version: 2.4.15
 Summary: pybaseutils
 Home-page: https://github.com/PanJinquan/base-utils
 Author: PanJinquan

{pybaseutils-2.4.11 → pybaseutils-2.4.15}/pybaseutils/__init__.py

@@ -8,4 +8,4 @@
 # --------------------------------------------------------
 """
 
-__version__ = '2.4.11'
+__version__ = '2.4.15'

pybaseutils-2.4.15/pybaseutils/cvutils/camera.py

@@ -0,0 +1,146 @@
+import cv2
+import time
+import numpy as np
+import subprocess
+import json
+import numbers
+
+
+def get_video_size(video, size=()):
+    """获取视频原始分辨率"""
+    if isinstance(video, int):  # TODO 摄像头
+        video = f"/dev/video{video}"
+    if size:
+        cmd = ['ffprobe', '-v', 'quiet',
+               '-video_size', f"{size[0]}x{size[1]}",  # 视频分辨率
+               '-print_format', 'json',
+               '-show_streams', video]
+    else:
+        cmd = ['ffprobe', '-v', 'quiet',
+               '-print_format', 'json',
+               '-show_streams', video]
+    result = subprocess.check_output(cmd).decode()
+    info = json.loads(result)
+    width = None
+    height = None
+    for stream in info['streams']:
+        if stream['codec_type'] == 'video':
+            width, height = stream['width'], stream['height']
+    print("video={},width:{}, height:{}".format(video, width, height))
+    return width, height
+
+
+class CameraCapture(object):
+    def __init__(self, video: str or int = 0, fps=30, size=(1920, 1080), scale=1.0):
+        """
+        查询视频设备分辨率： ffmpeg -f v4l2 -list_formats all -i /dev/video0
+        常见的视频分辨率  ： 1920x1080 1280x720 640x480 352x288 320x240 176x144 160x120
+        :param video: 视频设备路径或索引（如 0 或 "/dev/video0"）
+        :param fps: 视频帧率
+        :param size: 视频分辨率 (宽, 高)，(1280,720),(1920,1080)
+        :param scale: 视频缩放比例
+        """
+        self.fps = fps
+        self.stopped = False
+        self.ssize = get_video_size(video, size=size)
+        self.dsize = self.ssize
+        if isinstance(video, int):  # TODO 摄像头
+            video = f"/dev/video{video}"
+            self.video_size = f"{self.ssize[0]}x{self.ssize[1]}"  # 视频文件没有这个参数
+        else:  # TODO 视频文件
+            self.video_size = None
+        if size:
+            self.dsize = (int(size[0] * scale), int(size[1] * scale))
+        else:
+            self.dsize = (int(self.dsize[0] * scale), int(self.dsize[1] * scale))
+        vf = f'scale={self.dsize[0]}:{self.dsize[1]}'
+        # TODO FFmpeg 命令
+        # -re: 以原生帧率读取（模拟直播流）
+        # -fflags nobuffer: 关键！禁用缓冲区
+        # -flags low_delay: 关键！低延迟模式
+        # -probesize 32: 减小探测包大小，加快启动
+        # -pix_fmt bgr24: 直接输出 BGR 格式，方便 OpenCV/Numpy 使用，避免后续转换
+        if self.video_size:  # TODO 如果是摄像头
+            command = [
+                'ffmpeg',
+                '-re',  # 按帧率读取
+                '-fflags', 'nobuffer',  # 无缓冲
+                '-flags', 'low_delay',  # 低延迟
+                '-probesize', '32',  # 快速探测
+                '-video_size', f"{self.ssize[0]}x{self.ssize[1]}",  # 视频分辨率
+                '-i', video,  # 输入设备
+                '-f', 'rawvideo',  # 输出原始视频流
+                '-pix_fmt', 'bgr24',  # 像素格式 BGR
+                '-vf', vf,  # 分辨率
+                '-r', str(fps),  # 帧率
+                '-'  # 输出到 stdout
+            ]
+        else:
+            command = [
+                'ffmpeg',
+                '-re',  # 按帧率读取
+                '-fflags', 'nobuffer',  # 无缓冲
+                '-flags', 'low_delay',  # 低延迟
+                '-probesize', '32',  # 快速探测
+                '-i', video,  # 输入设备
+                '-f', 'rawvideo',  # 输出原始视频流
+                '-pix_fmt', 'bgr24',  # 像素格式 BGR
+                '-vf', vf,  # 分辨率
+                '-r', str(fps),  # 帧率
+                '-'  # 输出到 stdout
+            ]
+        # 启动进程
+        print(f"command: {' '.join(command)}")
+        self.pipe = subprocess.Popen(command, stdout=subprocess.PIPE, bufsize=10 ** 8)
+
+    def read(self):
+        if self.stopped:
+            self.stop()
+            return False, None
+        # 读取一帧的数据量 (宽 * 高 * 3通道)
+        buf = self.pipe.stdout.read(self.dsize[0] * self.dsize[1] * 3)
+        if len(buf) != self.dsize[0] * self.dsize[1] * 3:
+            self.stop()
+            return False, None  # 读取失败或结束
+        # TODO 将字节流转换为numpy数组图像(bgr)
+        bgr = np.frombuffer(buf, dtype=np.uint8).reshape((self.dsize[1], self.dsize[0], 3))
+        return True, bgr
+
+    def release(self):
+        self.stop()
+
+    def stop(self):
+        self.stopped = True
+        time.sleep(0.1)
+        self.pipe.terminate()  # 终止进程
+        try:
+            outs, errs = self.pipe.communicate(timeout=2)
+        except subprocess.TimeoutExpired as err:
+            # 如果 communicate 也超时（极少见），强制杀死
+            self.pipe.kill()
+            self.pipe.communicate()  # 再次尝试清理
+            print(err)
+
+    def display(self, title="camera", delay=30):
+        while True:
+            t1 = time.time()
+            ret, frame = self.read()
+            t2 = time.time()
+            if not ret:
+                break
+            t21 = (t2 - t1) * 1000
+            print(f"image shape: {frame.shape},耗时: {t21:.3f}ms")
+            cv2.namedWindow(title, flags=cv2.WINDOW_NORMAL)
+            cv2.imshow(title, frame)
+            cv2.waitKey(delay)
+
+
+if __name__ == '__main__':
+    fps = 10
+    video = 0  # Windows 下可能是 0 或 "video=Integrated Webcam"
+    # video = "/home/PKing/Videos/video1.mp4"  # Windows 下可能是 0 或 "video=Integrated Webcam"
+    # video = "../../data/video/kunkun_cut.mp4"  # Windows 下可能是 0 或 "video=Integrated Webcam"
+    # cap = CameraCapture(video=video, size=(), scale=1.0, fps=fps)
+    cap = CameraCapture(video=video, size=(1920, 1080), scale=1.0, fps=fps)
+    # cap = CameraCapture(video=video, size=(), scale=2,fps=fps)
+    cap.display()

{pybaseutils-2.4.11 → pybaseutils-2.4.15}/pybaseutils/dataloader/base_dataset.py

@@ -384,6 +384,7 @@ def get_targets_overlap(obj_info1: dict, obj_info2: dict, key="boxes", keys=[],
     boxes2 = obj_info2[key]
     nums1 = len(boxes1)
     nums2 = len(boxes2)
+    if nums1==0 or nums2==0:return []
     ious = image_utils.get_boxes_iom(boxes1, boxes2) if use_iom else image_utils.get_boxes_iou(boxes1, boxes2)
     output = []
     for i in range(len(boxes1)):

{pybaseutils-2.4.11 → pybaseutils-2.4.15}/pybaseutils/file_utils.py

@@ -78,6 +78,9 @@ def get_time(format="p"):
         # time = datetime.strftime(datetime.now(), '%Y%m%d_%H%M%S_%f')  # 20200508_143059_379116
         time = datetime.now().strftime("%Y%m%d_%H%M%S_%f")
         time = time[:-2]
+    elif format.lower() == "d":  # 精确到微妙
+        # time = datetime.strftime(datetime.now(), '%Y%m%d_%H%M%S_%f')  # 20200508_143059_379116
+        time = datetime.now().strftime("%Y%m%d")
     elif format.lower() == "y":  # 2025-06-18 11:02:05
         # time = datetime.strftime(datetime.now(), "%Y-%m-%d %H:%M:%S")  #
         time = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
@@ -891,6 +894,27 @@ def get_sub_list(file_list, dirname: str):
     return sub_list
 
 
+def get_sub_info(image_dir, postfix=IMG_POSTFIX, sub=0):
+    """
+    获得目image_dir下每个子文件夹的数据，以dict数据返回
+    :param image_dir:
+    :param postfix: 后缀
+    :param sub: 是否去除根路径
+    :return:  {"sub1":[file1,file2,...],"sub2":[file1,file2,...],"sub3":...}
+    """
+    sub_names = get_sub_paths(image_dir)
+    data_info = defaultdict(list)
+    for name in sub_names:
+        files = get_files_list(os.path.join(image_dir, name), postfix=postfix)
+        if sub == 1:
+            files = get_sub_list(files, dirname=image_dir)
+        elif sub == -1:
+            files = get_sub_list(files, dirname=os.path.dirname(image_dir))
+        data_info[name].extend(files)
+    data_info = dict(data_info)
+    return data_info
+
+
 def split_train_test(file_list, ratio=0.2, shuffle=False):
     """
     划分训练集和测试集
@@ -1146,15 +1170,17 @@ def print_dict(dict_data, save_path):
 def get_pair_data(image_dir, pair_num=-1):
     """
     获得图片对数据
+    Uages:
+    pairs = file_utils.get_pair_data(image_dir, pair_num=-1)
+    file_utils.write_data(pairs_file, pairs)
     :param image_dir:
     :param pair_num:-1 表示所有对
-    :return:
+    :return: [file1, file2, 1] 1表示相同类别，0表示不同类别
     """
     max_nums = int(pair_num / 2)
     image_list = get_files_lists(image_dir)
     image_list = get_sub_list(image_list, dirname=image_dir)
     nums = len(image_list)
-    print("have {} images and {} combinations".format(nums, nums * (nums - 1) / 2))
     pairs = []
     for paths in itertools.combinations(image_list, 2):
         file1, file2 = paths
@@ -1170,6 +1196,7 @@ def get_pair_data(image_dir, pair_num=-1):
     pair1 = pairs[pairs[:, -1] == "1", :]
     nums1 = len(pair1)
     nums0 = len(pair0)
+    print("have {} images, pair0:{}, pair1:{}".format(nums, nums0, nums1))
     if pair_num < 0: max_nums = nums1
     if max_nums > nums1:
         raise Exception("pair_nums({}) must be less than num_pair1({})".format(max_nums, nums1))
@@ -1191,7 +1218,7 @@ def get_pair_files(data_root, out_root=None, image_sub="", label_sub="",
     :param image_sub:
     :param label_sub:
     :param label_postfix: label文件后缀，如txt,png,json等
-    :return:
+    :return:  [file1,file2]
     """
     image_dir = os.path.join(data_root, image_sub)
     label_dir = os.path.join(data_root, label_sub)
@@ -1201,11 +1228,11 @@ def get_pair_files(data_root, out_root=None, image_sub="", label_sub="",
     for i, image_name in tqdm(enumerate(file_list)):
         postfix = image_name.split(".")[-1]
         lable_name = image_name.replace(f".{postfix}", f".{label_postfix}")
-        image_file = os.path.join(image_dir, image_name)
-        lable_file = os.path.join(label_dir, lable_name)
-        if os.path.exists(image_file) and os.path.exists(lable_file):
-            image_file, lable_file = get_sub_list([image_file, lable_file], dirname=data_root)
-            pair_list.append([image_file, lable_file])
+        file1 = os.path.join(image_dir, image_name)
+        file2 = os.path.join(label_dir, lable_name)
+        if os.path.exists(file1) and os.path.exists(file2):
+            file1, file2 = get_sub_list([file1, file2], dirname=data_root)
+            pair_list.append([file1, file2])
     if out_root:
         filename = os.path.join(out_root, "file_list.txt")
         write_data(filename, pair_list, split=",", mode='w')
@@ -1214,10 +1241,11 @@ def get_pair_files(data_root, out_root=None, image_sub="", label_sub="",
 
 def read_pair_data(filename, split=True):
     """
-    read pair data,data:[image1.jpg image2.jpg 0]
+    读取图片对数据，data:[image1.jpg image2.jpg 0]
+    可以使用get_pair_data(image_dir, pair_num=-1)获得图片对数据
     :param filename:
     :param split:
-    :return:
+    :return: [pair1, pair2, label] 1表示相同类别，0表示不同类别
     """
     pair_list = read_data(filename)
     if split:
@@ -1225,12 +1253,12 @@ def read_pair_data(filename, split=True):
         pair1 = pair_list[:, :1].reshape(-1)
         pair2 = pair_list[:, 1:2].reshape(-1)
         # convert to 0/1
-        issames_data = np.asarray(pair_list[:, 2:3].reshape(-1), dtype=np.int)
-        issames_data = np.where(issames_data > 0, 1, 0)
+        label = np.asarray(pair_list[:, 2:3].reshape(-1), dtype=np.int32)
+        label = np.where(label > 0, 1, 0)
         pair1 = pair1.tolist()
         pair2 = pair2.tolist()
-        issames_data = issames_data.tolist()
-        return pair1, pair2, issames_data
+        label = label.tolist()
+        return pair1, pair2, label
     return pair_list
 
 
@@ -1579,7 +1607,6 @@ def zip_file(src, dst=None, s=None):
     return dst
 
 
-
 if __name__ == '__main__':
     from pybaseutils import time_utils
 

{pybaseutils-2.4.11 → pybaseutils-2.4.15}/pybaseutils/image_utils.py

@@ -300,7 +300,7 @@ def get_prewhiten_image(x):
     '''
     mean = np.mean(x)
     std = np.std(x)
-    std_adj = np.maximum(std, 1.0 / np.sqrt(x.size))
+    std_adj = np.maximum(std, 1.0 / np.sqrt(x.dsize))
     y = np.multiply(np.subtract(x, mean), 1 / std_adj)
     return y
 

{pybaseutils-2.4.11 → pybaseutils-2.4.15}/pybaseutils/metrics/average_meter.py

@@ -63,7 +63,7 @@ class MultiAverageMeter(object):
 
 def accuracy(output, target, topk=(1, 5)):
     maxk = max(topk)
-    batch_size = target.size(0)
+    batch_size = target.dsize(0)
     _, pred = output.topk(maxk, 1, True, True)
     pred = pred.t()
     correct = pred.eq(target.view(1, -1).expand_as(pred))

{pybaseutils-2.4.11 → pybaseutils-2.4.15}/pybaseutils/metrics/class_report.py

@@ -12,6 +12,26 @@ from sklearn import metrics
 from pybaseutils import pandas_utils
 
 
+def get_metrics(true_labels, pred_labels, keys=['accuracy', 'precision', 'recall', 'f1_score']):
+    """
+    计算分类指标
+    :param true_labels: 真实样本的标签
+    :param pred_labels: 预测的标签
+    :param keys: 计算的指标，默认计算accuracy, precision, recall, f1_score
+    :return: 分类指标
+    """
+    aprf = {}
+    if 'accuracy' in keys:
+        aprf['accuracy'] = metrics.accuracy_score(true_labels, pred_labels)
+    if 'precision' in keys:
+        aprf['precision'] = metrics.precision_score(true_labels, pred_labels, average=None)  # 每个类别
+    if 'recall' in keys:
+        aprf['recall'] = metrics.recall_score(true_labels, pred_labels, average=None)
+    if 'f1_score' in keys:
+        aprf['f1_score'] = metrics.f1_score(true_labels, pred_labels, average=None)
+    return aprf
+
+
 def plot_confusion_matrix(conf_matrix, labels_name, title, normalization=True):
     if normalization:
         conf_matrix = conf_matrix.astype('float') / conf_matrix.sum(axis=1)[:, np.newaxis]  # 归一化
@@ -67,10 +87,17 @@ def get_confusion_matrix(true_labels, pred_labels, target_names=None, filename=N
 def get_metrics_report(true_labels, pred_labels, target_names=None, labels=None,
                        output_dict=False, matrix=False, plot=False):
     """
+    Usage
+    -------
+    true_labels = ["A", "B", "A", "B"]
+    pred_labels = ["A", "B", "C", "B"]
+    out_result = get_metrics_report(true_labels, pred_labels)
+    -------
     true_labels = [0, 1, 2, 3, 4, 1]  # Y
     pred_labels = [0, 1, 1, 2, 2, 1]  # X
     target_names = ["A", "B", "C", "D", "E"]
-    out_result = get_metrics_report(true_labels, pred_labels, target_names=target_names, output_dict=False)
+    out_result = get_metrics_report(true_labels, pred_labels, target_names=target_names)
+    -------
     宏平均(macro avg)和微平均(micro avg)
     如果每个class的样本数量差不多,那么宏平均和微平均没有太大差异
     如果每个class的样本数量差异很大,而且你想:
@@ -84,32 +111,33 @@ def get_metrics_report(true_labels, pred_labels, target_names=None, labels=None,
     :param matrix: 是否绘制混淆矩阵
     :return:
     """
-    # if target_names is None:
-    #     target_names = list(set(pred_labels) | set(true_labels))
-    # else:
-    #     true_labels = [target_names[int(i)] for i in true_labels]
-    #     pred_labels = [target_names[int(i)] for i in pred_labels]
     result = metrics.classification_report(true_labels,
                                            pred_labels,
                                            labels=labels,
                                            digits=4,
                                            target_names=target_names,
-                                           output_dict=output_dict,
+                                           output_dict=True,
                                            zero_division=0)
     if output_dict:
         macro_avg = result["macro avg"]
-        accuracy = result["accuracy"]
         weighted_avg = result["weighted avg"]
-        out_result = {"macro_avg": macro_avg, "accuracy": accuracy, "weighted_avg": weighted_avg}
+        accuracy = result["accuracy"]
+        output = {"macro_avg": macro_avg, "accuracy": accuracy, "weighted_avg": weighted_avg}
         # pdf=pd.DataFrame.from_dict(result)
         # save_csv("classification_report.csv", pdf)
     else:
-        out_result = result
+        p = (result['macro avg']['precision'] + result['weighted avg']['precision']) / 2
+        r = (result['macro avg']['recall'] + result['weighted avg']['recall']) / 2
+        support = result['macro avg']['support']
+        result["accuracy"] = {'precision': p, 'recall': r, 'f1-score': result["accuracy"], 'support': support}
+        output = pandas_utils.dict2df(result)
+        output = output.round(4)  # 保留4位小数
+        output = output.to_markdown()
     if matrix:
         conf_matrix = get_confusion_matrix(true_labels, pred_labels, target_names=target_names,
                                            normalization=True, plot=plot, title="Confusion Matrix")
         # out_result["confusion_matrix"] = conf_matrix
-    return out_result
+    return output
 
 
 get_classification_report = get_metrics_report
@@ -148,11 +176,10 @@ def create_file_path(filename):
 if __name__ == "__main__":
     # true_labels = [0, 1, 2, 3, 3, 1, 1]  # Y
     # pred_labels = [1, 1, 2, 2, 2, 1, 0]  # X
-    true_labels = ["A", "B", "A", "B"]
-    pred_labels = ["A", "B", "A", "B"]
-    # target_names = ["A0", "B1", "C2", "D3"]
+    true_labels = ["A", "B", "A", "B", "C", "D"]
+    pred_labels = ["A", "D", "C", "B", "C", "A"]
     target_names = None
     confuse_file = "./confuse.csv"
-    out_result = get_classification_report(true_labels, pred_labels, target_names=target_names, output_dict=False,
-                                           matrix=True, plot=True)
-    print(out_result)
+    result = get_metrics_report(true_labels, pred_labels, target_names=target_names, output_dict=False,
+                                matrix=True, plot=False)
+    print(result)

{pybaseutils-2.4.11 → pybaseutils-2.4.15}/pybaseutils/numpy_utils.py

@@ -489,13 +489,13 @@ def L2_loss(y_true, y_pre):
 def mean_squared_error(y_true, y_pre):
     """MSE(Mean Squared Error)平均平方误差(L2)"""
     l2 = np.sum(np.square(y_true - y_pre))
-    return l2 / y_true.size
+    return l2 / y_true.dsize
 
 
 def mean_absolute_error(y_true, y_pre):
     """MAE(Mean Absolute Error)平均绝对差值(L1),也等于MAD(Mean Absolute Difference)"""
     l1 = np.sum(np.abs(y_true - y_pre))
-    return l1 / y_true.size
+    return l1 / y_true.dsize
 
 
 def mean(data):

{pybaseutils-2.4.11 → pybaseutils-2.4.15}/pybaseutils.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pybaseutils
-Version: 2.4.11
+Version: 2.4.15
 Summary: pybaseutils
 Home-page: https://github.com/PanJinquan/base-utils
 Author: PanJinquan

{pybaseutils-2.4.11 → pybaseutils-2.4.15}/test_py/aije/action_dataset.py

@@ -47,7 +47,7 @@ def save_object_crops(data_info, out_dir, class_name=None, scale=[], square=Fals
                                                  thickness=2, fontScale=0.8, drawType="chinese")
         image_utils.cv_show_image("image", m, use_rgb=False, delay=0)
     for i, img in enumerate(crops):
-        if img.size < min_size: continue
+        if img.dsize < min_size: continue
         name = class_name[int(labels[i])] if class_name else labels[i]
         if out_dir:
             file_name = "{}_{:0=4d}_{}.jpg".format(image_id, i, flag) if flag else "{}_{:0=4d}.jpg".format(image_id, i)

{pybaseutils-2.4.11 → pybaseutils-2.4.15}/test_py/aije/demo_labelme_crop_aije.py

@@ -49,7 +49,7 @@ def save_object_crops(data_info, out_dir, class_name=None, target_name=None, sca
                                                  thickness=2, fontScale=0.8, drawType='chinese')
         image_utils.cv_show_image('image', m, use_rgb=False, delay=0)
     for i, img in enumerate(crops):
-        if img.size < min_size: continue
+        if img.dsize < min_size: continue
         name = class_name[int(labels[i])] if class_name else labels[i]
         if out_dir:
             file_name = '{}_{:0=4d}_{}.jpg'.format(image_id, i, flag) if flag else '{}_{:0=4d}.jpg'.format(image_id, i)

{pybaseutils-2.4.11 → pybaseutils-2.4.15}/test_py/aije/demo_labelme_crop_ailt.py

@@ -49,7 +49,7 @@ def save_object_crops(data_info, out_dir, class_name=None, target_name=None, sca
                                                  thickness=2, fontScale=0.8, drawType='chinese')
         image_utils.cv_show_image('image', m, use_rgb=False, delay=0)
     for i, img in enumerate(crops):
-        if img.size < min_size: continue
+        if img.dsize < min_size: continue
         name = class_name[int(labels[i])] if class_name else labels[i]
         if out_dir:
             file_name = '{}_{:0=4d}_{}.jpg'.format(image_id, i, flag) if flag else '{}_{:0=4d}.jpg'.format(image_id, i)
@@ -106,17 +106,15 @@ if __name__ == '__main__':
                    '胶枪灯', '胶枪灯红色', '胶枪灯绿色',
                    'PDA', 'PDA*', '清洁布', '手与*接触', '电视柜台面', '*标签'
                    ]
-    # '*长条标签'
     datasets = [
-        '/home/PKing/nasdata/dataset-dmai/AILT/ailt-det/dataset-20260227/images',
-        '/home/PKing/nasdata/dataset-dmai/AILT/ailt-det/dataset-20260211/images',
+        '/home/PKing/nasdata/dataset-dmai/AILT/ailt-det/dataset-20260420/images',
     ]
     # datasets = "/home/PKing/nasdata/dataset-dmai/AILT/ailt-det/文件列表.txt"
-    datasets = file_utils.read_data(datasets, split=None)
+    # datasets = file_utils.read_data(datasets, split=None)
     # dataroot = '/home/PKing/nasdata/dataset-dmai/AIJE/dataset/aije-v2-det'
     # datasets = get_sub_dataset(dataroot)
     class_name = []
-    datasets = sorted(datasets)
+    # datasets = sorted(datasets)
     for anno_dir in datasets:
         print(anno_dir)
         assert os.path.exists(anno_dir), anno_dir

{pybaseutils-2.4.11 → pybaseutils-2.4.15}/test_py/aije/demo_labelme_shock.py

@@ -46,7 +46,7 @@ def save_object_crops(data_info, item_list, out_dir, class_name=None, scale=[],
         else:
             crops = image_utils.get_bboxes_crop(image, bboxes)
         img = crops[0]
-        if img.size < min_size: continue
+        if img.dsize < min_size: continue
         name = "_".join(labels + list(set(attrib_name)))
         if out_dir:
             file_name = "{}_{:0=4d}_{}.jpg".format(image_id, i, flag) if flag else "{}_{:0=4d}.jpg".format(image_id,

{pybaseutils-2.4.11 → pybaseutils-2.4.15}/test_py/demo2.py

@@ -61,6 +61,20 @@ def video_iterator(video):
 
 
 if __name__ == '__main__':
-    x = 123
-    x = None
-    print(f"{str(x):7}")
+    from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score
+
+    true_labels = ["A", "B", "A", "B", "C"]
+    pred_labels = ["A", "B", "C", "B", "C"]
+
+    accuracy = accuracy_score(true_labels, pred_labels)
+    precision = precision_score(true_labels, pred_labels, average=None)  # 每个类别
+    recall = recall_score(true_labels, pred_labels, average=None)
+    f1_score = f1_score(true_labels, pred_labels, average=None)
+
+    print(f"准确率: {accuracy}")
+    print(f"精确率: {dict(zip(set(true_labels), precision))}")
+    print(f"召回率: {dict(zip(set(true_labels), recall))}")
+    print(f"F1值: {dict(zip(set(true_labels), f1_score))}")
+
+
+

{pybaseutils-2.4.11 → pybaseutils-2.4.15}/test_py/demo_labelme_crop.py

@@ -48,7 +48,7 @@ def save_object_crops(data_info, out_dir, class_name=None, scale=[], square=Fals
                                                  thickness=2, fontScale=0.8, drawType="chinese")
         image_utils.cv_show_image("image", m, use_rgb=False, delay=0)
     for i, img in enumerate(crops):
-        if img.size < min_size: continue
+        if img.dsize < min_size: continue
         name = class_name[int(labels[i])] if class_name else labels[i]
         if out_dir:
             file_name = "{}_{:0=4d}_{}.jpg".format(image_id, i, flag) if flag else "{}_{:0=4d}.jpg".format(image_id, i)

{pybaseutils-2.4.11 → pybaseutils-2.4.15}/test_py/single_class/demo.py

@@ -89,7 +89,7 @@ class LabelSmoothingCrossEntropy(nn.Module):
         self.reduction = reduction
 
     def forward(self, output, target):
-        c = output.size()[-1]
+        c = output.dsize()[-1]
         log_preds = F.log_softmax(output, dim=-1)
         if self.reduction == 'sum':
             loss = -log_preds.sum()

{pybaseutils-2.4.11 → pybaseutils-2.4.15}/test_py/slowfastnet.py

@@ -111,7 +111,7 @@ class SlowFast(nn.Module):
         x = torch.cat([x, lateral[3]],dim=1)
         x = self.slow_res5(x)
         x = nn.AdaptiveAvgPool3d(1)(x)
-        x = x.view(-1, x.size(1))
+        x = x.view(-1, x.dsize(1))
         return x
 
     def FastPath(self, input):
@@ -137,7 +137,7 @@ class SlowFast(nn.Module):
 
         res5 = self.fast_res5(res4)
         x = nn.AdaptiveAvgPool3d(1)(res5)
-        x = x.view(-1, x.size(1))
+        x = x.view(-1, x.dsize(1))
 
         return x, lateral
 
@@ -214,4 +214,4 @@ if __name__ == "__main__":
     input_tensor = torch.autograd.Variable(torch.rand(1, 3, 64, 224, 224))
     model = resnet50(class_num=num_classes)
     output = model(input_tensor)
-    print(output.size())
+    print(output.dsize())