json 可视化 2025 coco json

目录

查看各种类型;

[coco json可视化](#coco json可视化)


查看各种类型;

python 复制代码
import glob
import json
import os
import time

import numpy as np
import cv2
from PIL import Image, ImageDraw, ImageFont
from natsort import natsorted

def draw_shapes(json_path):
    # 读取 JSON
    with open(json_path, 'r', encoding='utf-8') as f:
        data = json.load(f)

    img_path = os.path.join(os.path.dirname(json_path), data["imagePath"])
    img = cv2.imread(img_path)
    if img is None:
        print(f"图片未找到: {img_path}")
        return

    shapes = data["shapes"]

    for shape in shapes:
        label = shape["label"]
        shape_type = shape["shape_type"]
        points = np.array(shape["points"], dtype=np.int32)

        color = (0, 255, 0)  # 默认绿色
        if shape_type == "ellipse":
            # points: [[cx, cy], [w, h]]
            cx, cy = map(int, points[0])
            w, h = map(int, points[1])
            cv2.ellipse(img, (cx, cy), (int(w / 2), int(h / 2)), 0, 0, 360, color, 2)

        elif shape_type in ["polygon", "contour"]:
            cv2.polylines(img, [points], isClosed=True, color=color, thickness=2)

        elif shape_type == "points":
            for (x, y) in points:
                cv2.circle(img, (int(x), int(y)), 3, (0, 0, 255), -1)

        elif shape_type == "rectangle":
            x1, y1 = points[0]
            x2, y2 = points[1]
            cv2.rectangle(img, (int(x1), int(y1)), (int(x2), int(y2)), color, 2)

        else:
            print(f"⚠️ 未知类型: {shape_type}")
            continue

        # 绘制文字标签
        cv2.putText(img, label, (int(points[0][0]), int(points[0][1]) - 5),
                    cv2.FONT_HERSHEY_SIMPLEX, 0.6, (0, 255, 255), 2)

    return img

if __name__ == '__main__':

    font = ImageFont.truetype("simhei.ttf", 22, encoding="utf-8")

    img_to =r'D:\data\tmp\data_similar\0\ellipse\1010_1905_0/'
    img_to =r'D:\data\tmp\data_similar\0\box_frame\1010_1938_0/'
    img_to =r'D:\data\tmp\data_similar\0\box_frame\1011_2037_0/'
    img_to =r'D:/data/tmp/data_similar/0/box_frame/1011_2037_0/shear_y/rotate/'
    img_to =r'D:/data/tmp/data_similar/0/box_frame/1011_2037_0/rotate/'
    img_to =r'D:\data\course_1023\jiezhi\20251023_1738_part001_seg/'
    img_to =r'C:\Users\ChanJing-01\Desktop\tmp\1024/'
    img_to =r'D:\data\Preview_10\20251021-185957_seg/'

    json_paths =glob.glob(img_to + "/*.json")

    json_paths=natsorted(json_paths,reverse=True)

    for json_path in json_paths:

        img = draw_shapes(json_path)

        cv2.imshow("Annotation Viewer", img)
        cv2.waitKey(0)
    cv2.destroyAllWindows()

coco json可视化

python 复制代码
import glob
import json
import os
import time

import numpy as np
import cv2
from PIL import Image, ImageDraw, ImageFont
from natsort import natsorted


def iou(box1, box2):
    '''
    box:[x1, y1, x2, y2]
    '''
    in_h = min(box1[3], box2[3]) - max(box1[1], box2[1])
    in_w = min(box1[2], box2[2]) - max(box1[0], box2[0])
    inner = 0 if in_h <0 or in_w <0 else in_h *in_w
    union = (box1[2] - box1[0]) * (box1[3] - box1[1]) + \
            (box2[2] - box2[0]) * (box2[3] - box2[1]) - inner
    iou = inner / union
    return iou


def iou_over(box1, box2):
    '''
    box:[x1, y1, x2, y2] 相交区域 最小的box的占比
    '''
    in_h = min(box1[3], box2[3]) - max(box1[1], box2[1])
    in_w = min(box1[2], box2[2]) - max(box1[0], box2[0])
    inner = 0 if in_h <0 or in_w <0 else in_h *in_w
    union = min((box1[2] - box1[0]) * (box1[3] - box1[1]), (box2[2] - box2[0]) * (box2[3] - box2[1]))
    iou = inner / union
    return iou

def batch_IoU(bbox, gt):
    """
    :param bbox: (n, 4)
    :param gt: (m, 4)
    :return: (n, m)
    numpy 广播机制 从后向前对齐。 维度为1 的可以重复等价为任意维度
    eg: (4,3,2)   (3,2)  (3,2)会扩充为(4,3,2)
        (4,1,2)   (3,2) (4,1,2) 扩充为(4, 3, 2)  (3, 2)扩充为(4, 3,2) 扩充的方法为重复
    广播会在numpy的函数 如sum, maximun等函数中进行
    pytorch同理。
    扩充维度的方法:
    eg: a  a.shape: (3,2)  a[:, None, :] a.shape: (3, 1, 2) None 对应的维度相当于newaxis
    """
    lt = np.maximum(bbox[:, None, :2], gt[:, :2])  # left_top (x, y)
    rb = np.minimum(bbox[:, None, 2:], gt[:, 2:])  # right_bottom (x, y)
    wh = np.maximum(rb - lt , 0)                # inter_area (w, h)
    inter_areas = wh[:, :, 0] * wh[:, :, 1]        # shape: (n, m)
    box_areas = (bbox[:, 2] - bbox[:, 0] ) * (bbox[:, 3] - bbox[:, 1] )
    gt_areas = (gt[:, 2] - gt[:, 0] ) * (gt[:, 3] - gt[:, 1] )
    IoU = inter_areas / (box_areas[:, None] + gt_areas - inter_areas)
    return IoU

class MyEncoder(json.JSONEncoder):
    def default(self, obj):
        if isinstance(obj, np.integer):
            return int(obj)
        elif isinstance(obj, np.floating):
            return float(obj)
        elif isinstance(obj, np.ndarray):
            return obj.tolist()
        else:
            return super(MyEncoder, self).default(obj)

if __name__ == '__main__':

    font = ImageFont.truetype("simhei.ttf", 22, encoding="utf-8")

    img_to =r'D:\data\quexian\det_data\det_0915\rotated_results/'
    img_to =r'D:\data\tmp\data_label\video2/'
    img_to =r'D:\data\tmp\data_similar\0928_1747_1/'
    img_to =r'D:\data\tmp\data_similar\0\string\0929_1812_0\stretch/'

    from_labels =glob.glob(img_to + "/*.json")

    from_labels=natsorted(from_labels,reverse=True)

    for f_label_file in from_labels:
        base_name= os.path.basename(f_label_file)
        to_val = json.load(open(f_label_file, 'r',encoding='utf-8'))
        imagePath = to_val['imagePath']
        img_path= img_to + imagePath
        print(img_path)
        if "shapes" not in to_val:
            continue
        to_shapes =to_val['shapes']

        if not os.path.exists(img_path):
            img_path=f_label_file.replace('.json', '.jpg')
        frame=cv2.imread(img_path)
        if len(to_shapes)<1:
            print('none',img_path)
        for to_shape in to_shapes:

            if to_shape["shape_type"]== "rectangle":

                print('label',to_shape['label'])
                box2 = to_shape['points'][0] + to_shape['points'][1]

                points = np.array(to_shape["points"])
                xmin = (min(points[:, 0]) if min(points[:, 0]) > 0 else 0)
                xmax = (max(points[:, 0]) if max(points[:, 0]) > 0 else 0)
                ymin = (min(points[:, 1]) if min(points[:, 1]) > 0 else 0)
                ymax = (max(points[:, 1]) if max(points[:, 1]) > 0 else 0)

                cv2.rectangle(frame, (int(xmin), int(ymin)), (int(xmax), int(ymax)), (0, 0, 255), 1)

                start = time.time()
                cv2_im = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
                pil_im = Image.fromarray(cv2_im)
                draw = ImageDraw.Draw(pil_im)
                draw.text((int(xmin), int(ymin)-10), to_shape['label'], (255, 0, 0), font=font)
                frame = cv2.cvtColor(np.array(pil_im), cv2.COLOR_RGB2BGR)
            # cv2.rectangle(img, (int(to_shape['points'][0][0]), int(to_shape['points'][0][1])), (int(to_shape['points'][1][0]), int(to_shape['points'][1][1])), (0, 0, 255), 2)

        if np.prod(frame.shape[:2]) > 1000 * 1300:
            x_scale = np.sqrt(1000 * 1200 / np.prod(frame.shape[:2]))

            new_width = int(frame.shape[1] * x_scale)
            new_height = int(frame.shape[0] * x_scale)

            # 确保宽高都是偶数
            if new_width % 2 != 0:
                new_width += 1
            if new_height % 2 != 0:
                new_height += 1
            frame = cv2.resize(frame, (new_width, new_height), interpolation=cv2.INTER_AREA)
        cv2.imshow("image" ,frame)
        cv2.waitKey(0)
相关推荐
Wellijohn9 小时前
我把公司内部的 12 个 CRUD 接口交给了 AI Agent,然后省出了两周
人工智能·python·mcp
2601_9679801110 小时前
python pytest怎样测试代码
python·pytest·测试·文档·断言
伞伞悦读10 小时前
【第92期】逻辑回归的分数概率与阈值的实验与排查
人工智能·python·深度学习·机器学习
ch62311 小时前
🚀 从零构建本地文档智能审查 Agent:Vibe Coding 实战指南
python
摇滚侠11 小时前
《On Java 中文版 基础卷》阅读笔记 接口 18
java·笔记·python
Thanks_ks12 小时前
AI Agent 开发入门:用 Function Calling 让大模型学会 “动手干活”,附订单查询机器人实战
python·ai·大模型·openai·agent·通义千问·mcp
专注仿真12 小时前
智能体训练之模仿学习:GAIL 算法详解与实战
python·算法·仿真·多智能体强化学习·afsim智能体训练
databook13 小时前
处理一批 PDF,怎么算真的处理完了
后端·python
承渊政道15 小时前
【Python编程快速上手—让繁琐工作自动化】(列表:把数据排成队,从下标取值到深拷贝)
开发语言·python·pycharm·自动化·列表
大胖爱学习_MJ16 小时前
Agent 如何安全地拿数据库密码?基于 HashiCorp Vault 的动态凭证下发
经验分享·python