目录
[coco json可视化](#coco json可视化)
查看各种类型；
python
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
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)