使用Plotly的px.imshow绘制多图像子图报形状错误如何解决？

• 触发报错的核心是两个用法问题：
  1. px.imshow()返回的是完整的Figure对象，不是可被add_trace()接收的单条trace对象，不能直接和make_subplots搭配使用
  2. 你传入了包含5张单通道图的三维数组(5, 1440, 1920)，不符合px.imshow()默认仅接收单张2D灰度/3通道RGB/4通道RGBA图的要求

有两种常用实现方式，按需选择即可：

方案1：用make_subplots逐张添加图像 #

使用plotly.graph_objects的Image类作为trace添加到子图中，代码修改如下：

from skimage.morphology import binary_closing, binary_dilation, binary_erosion, binary_opening, selem
import plotly.graph_objects as go
from plotly.subplots import make_subplots
import cv2
import os


file = "./../../...zip"
image_file = "./../../images"

# 注意你原代码漏了img的读取逻辑，自行替换为实际图像路径即可
img = cv2.imread("你的实际图像路径", 0) # 按灰度图读取
img = cv2.convertScaleAbs(img)
img = cv2.medianBlur(img, 5)

# use a disk of radius 3
selem = selem.disk(3)

# opening and closing
open_img = binary_opening(img, selem)
close_img = binary_closing(img, selem)

# erosion and dilation
eroded_img = binary_erosion(img, selem)
dilated_img = binary_dilation(img, selem)

# 定义子图标题
subplot_titles = ["原图", "开运算", "闭运算", "腐蚀", "膨胀"]
# 创建3行2列子图
fig = make_subplots(rows=3, cols=2, subplot_titles=subplot_titles)

# 逐张添加图像到对应位置
fig.add_trace(go.Image(z=img, colorscale='gray'), row=1, col=1)
fig.add_trace(go.Image(z=open_img, colorscale='gray'), row=1, col=2)
fig.add_trace(go.Image(z=close_img, colorscale='gray'), row=2, col=1)
fig.add_trace(go.Image(z=eroded_img, colorscale='gray'), row=2, col=2)
fig.add_trace(go.Image(z=dilated_img, colorscale='gray'), row=3, col=1)

# 调整布局后显示
fig.update_layout(height=900, width=1200, title_text="形态学操作效果对比")
fig.show()

方案2：直接用px.imshow的facet_col参数自动生成分栏子图 #

不用手动创建子图，直接传入三维数组，指定分栏参数即可，代码更简洁：

from skimage.morphology import binary_closing, binary_dilation, binary_erosion, binary_opening, selem
import plotly.express as px
import cv2
import os


file = "./../../...zip"
image_file = "./../../images"

# 自行替换为实际图像路径
img = cv2.imread("你的实际图像路径", 0)
img = cv2.convertScaleAbs(img)
img = cv2.medianBlur(img, 5)

# use a disk of radius 3
selem = selem.disk(3)

# opening and closing
open_img = binary_opening(img, selem)
close_img = binary_closing(img, selem)

# erosion and dilation
eroded_img = binary_erosion(img, selem)
dilated_img = binary_dilation(img, selem)

# 把所有图拼成列表，传入px.imshow，指定facet_col按第0维分栏
fig = px.imshow(
    [img, open_img, close_img, eroded_img, dilated_img],
    facet_col=0, # 沿第0个维度拆分，每一组作为一个子图
    facet_col_wrap=2, # 每行最多放2张子图
    color_continuous_scale='gray',
    title="形态学操作效果对比"
)
# 替换默认子图标题为自定义内容
for i, title in enumerate(["原图", "开运算", "闭运算", "腐蚀", "膨胀"]):
    fig.layout.annotations[i].text = title

fig.update_layout(height=900, width=1200)
fig.show()

内容的提问来源于stack exchange，提问作者melololo