Core ML模型预测报错：imagePath需为PIL.Image.Image类型求解决方案

问题 #

我有一个针对1920x1080分辨率车牌训练的.mlmodel模型，测试视频分辨率同样为1920x1080。运行代码时出现如下错误：

typeerror: Image input, 'imagePath' must be of type PIL.Image.Image in the input dict

报错代码行：

# Predict using Core ML model
output = model.predict({"imagePath": temp.name, "iouThreshold": 0.5, "confidenceThreshold": 0.5})

完整代码如下：

import cv2

model = coremltools.models.MLModel('NeuralVision.mlmodel')

video_path = 'anpr_video.mp4'
cap = cv2.VideoCapture(video_path)

# Desired size for the model input
new_width = 1920
new_height = 1080

# Process the video frame by frame
while cap.isOpened():
    ret, frame = cap.read()

    # Break the loop if we have reached the end of the video
    if not ret:
        break

    # Convert the frame to a PIL Image
    image = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))

    # Resize the image to the desired size
    resized_image = image.resize((new_width, new_height), Image.LANCZOS)

    # Save the resized image as a temporary file
    with tempfile.NamedTemporaryFile(suffix=".png", delete=False) as temp:
        resized_image.save(temp.name)

        # Predict using Core ML model
        output = model.predict({"imagePath": temp.name, "iouThreshold": 0.5, "confidenceThreshold": 0.5})

    # Print the output
    print(output)

    # Display the frame
    cv2.imshow('Frame', frame)

    # Break the loop if the user presses the 'q' key
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

# Release the video capture object and close all windows
cap.release()
cv2.destroyAllWindows()

解决方法 #

错误原因清晰：你的Core ML模型要求imagePath输入是PIL.Image.Image类型，但你传入的是临时文件路径字符串，触发了类型错误。直接传入处理好的PIL图像即可，无需存临时文件，修改步骤如下：

• 删除临时文件相关代码：不需要将PIL图像存为文件再读取，直接使用内存中的图像对象
• 修改predict的输入参数：把temp.name替换为处理好的resized_image
• 补上缺失的模块导入：原代码缺少from PIL import Image、import tempfile和import coremltools语句，必须添加

修改后的完整代码：

import cv2
from PIL import Image
import tempfile
import coremltools

model = coremltools.models.MLModel('NeuralVision.mlmodel')

video_path = 'anpr_video.mp4'
cap = cv2.VideoCapture(video_path)

# Desired size for the model input
new_width = 1920
new_height = 1080

# Process the video frame by frame
while cap.isOpened():
    ret, frame = cap.read()

    # Break the loop if we have reached the end of the video
    if not ret:
        break

    # Convert the frame to a PIL Image
    image = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))

    # Resize the image to the desired size
    resized_image = image.resize((new_width, new_height), Image.LANCZOS)

    # Predict using Core ML model - 直接传入PIL图像对象
    output = model.predict({"imagePath": resized_image, "iouThreshold": 0.5, "confidenceThreshold": 0.5})

    # Print the output
    print(output)

    # Display the frame
    cv2.imshow('Frame', frame)

    # Break the loop if the user presses the 'q' key
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

# Release the video capture object and close all windows
cap.release()
cv2.destroyAllWindows()

内容的提问来源于stack exchange，提问作者Elise siera