Loading Yolov9 Model Trained on Custom Dataset: AttributeError: ‘str’ object has no attribute ‘shape’

Welcome to this comprehensive guide on loading a Yolov9 model trained on a custom dataset, where we’ll tackle the common error “AttributeError: ‘str’ object has no attribute ‘shape'”. This error can be frustrating, but don’t worry, we’ve got you covered. By the end of this article, you’ll be loading your custom Yolov9 model like a pro!

What is Yolov9?

Yolov9 is an object detection algorithm that has gained immense popularity in recent years due to its high accuracy and real-time performance. It’s an improved version of the YOLO (You Only Look Once) family, which has been widely used in various computer vision applications, including self-driving cars, surveillance systems, and more.

What is a Custom Dataset?

A custom dataset is a collection of images or data that you’ve gathered and annotated specifically for your project. This dataset is unique to your needs and can be used to train machine learning models, including Yolov9. By training a model on a custom dataset, you can achieve better performance and accuracy on your specific task.

The Error: AttributeError: ‘str’ object has no attribute ‘shape’

So, you’ve trained your Yolov9 model on your custom dataset, and now you’re trying to load it, but you’re getting the error “AttributeError: ‘str’ object has no attribute ‘shape'”. This error occurs when you’re trying to access the shape attribute of a string object, which doesn’t exist.

This error can occur due to various reasons, including:

• Incorrect model loading
• Invalid dataset format
• Mismatched data types
• Incorrect annotation format

Step-by-Step Guide to Loading Yolov9 Model Trained on Custom Dataset

Let’s dive into the step-by-step guide to loading your Yolov9 model trained on a custom dataset:

Step 1: Prepare Your Custom Dataset

Before loading the model, make sure your custom dataset is properly formatted and annotated. Your dataset should consist of:

• Images (JPEG or PNG)
• Annotation files (XML or CSV)

Ensure that your annotation files are in the correct format, with the following information:

Annotation Format	Example
XML (Pascal VOC)	<annotation> <folder>Folder Name</folder> <filename>Image Name</filename> <path>Image Path</path> <object> <name>Object Class</name> <bndbox> <xmin>xmin</xmin> <ymin>ymin</ymin> <xmax>xmax</xmax> <ymax>ymax</ymax> </bndbox> </object> </annotation>
CSV	Image Path,Object Class,xmin,ymin,xmax,ymax

Step 2: Install Required Libraries

Make sure you have the required libraries installed:

pip install opencv-python
pip install yolov9

Step 3: Load the Yolov9 Model

Load the Yolov9 model using the following code:

import cv2

# Load the Yolov9 model
net = cv2.dnn.readNet("yolov9s.cfg", "yolov9s.weights")

Step 4: Load the Custom Dataset

Load your custom dataset using the following code:

import pandas as pd

# Load the annotation file (XML or CSV)
if annotation_format == "xml":
    annotation_file = "annotations.xml"
    tree = ET.parse(annotation_file)
    root = tree.getroot()
    annotations = []
    for obj in root.findall("object"):
        annotations.append({
            "image_id": obj.find("filename").text,
            "class_id": obj.find("name").text,
            "xmin": int(obj.find("bndbox/xmin").text),
            "ymin": int(obj.find("bndbox/ymin").text),
            "xmax": int(obj.find("bndbox/xmax").text),
            "ymax": int(obj.find("bndbox/ymax").text)
        })
elif annotation_format == "csv":
    annotation_file = "annotations.csv"
    annotations = pd.read_csv(annotation_file)

Step 5: Preprocess the Data

Preprocess the data by resizing the images and converting them to the required format:

import numpy as np

# Resize the images
image_size = (416, 416)
for annotation in annotations:
    image = cv2.imread(annotation["image_id"])
    image = cv2.resize(image, image_size)
    annotation["image"] = image

# Convert the data to the required format
data = []
for annotation in annotations:
    image = annotation["image"]
    class_id = annotation["class_id"]
    xmin = annotation["xmin"]
    ymin = annotation["ymin"]
    xmax = annotation["xmax"]
    ymax = annotation["ymax"]
    data.append({
        "image": image,
        "class_id": class_id,
        "bbox": [xmin, ymin, xmax, ymax]
    })

Step 6: Load the Model and Make Predictions

Load the model and make predictions using the following code:

# Set the model input
blob = cv2.dnn.blobFromImage(image, 1/255, (416, 416), swapRB=True, crop=False)

# Set the model input
net.setInput(blob)

# Make predictions
outputs = net.forward(net.getUnconnectedOutLayersNames())

# Get the detections
detections = []
for output in outputs:
    for detection in output:
        scores = detection[5:]
        class_id = np.argmax(scores)
        confidence = scores[class_id]
        if confidence > 0.5 and class_id == 0:
            x, y, w, h = detection[0:4] * 416
            detections.append({
                "class_id": class_id,
                "confidence": confidence,
                "bbox": [x, y, w, h]
            })

# Display the detections
for detection in detections:
    x, y, w, h = detection["bbox"]
    cv2.rectangle(image, (int(x), int(y)), (int(x+w), int(y+h)), (0, 255, 0), 2)
    cv2.putText(image, f"{detection['class_id']} {detection['confidence']:.2f}", (int(x), int(y-10)), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0), 2)

Common Mistakes and Solutions

Here are some common mistakes and solutions to help you troubleshoot the error:

Mistake 1: Incorrect Model Loading

Solution: Ensure that the model configuration file (yolov9s.cfg) and weights file (yolov9s.weights) are in the correct location and format.

Mistake 2: Invalid Dataset Format

Solution: Verify that your dataset is in the correct format, with the correct annotation files and image paths.

Mistake 3: Mismatched Data Types

Solution: Ensure that the data types of the input data and model inputs match.

Mistake 4: Incorrect Annotation Format

Solution: Verify that the annotation file format is correct, with the correct class names and bounding box coordinates.

Conclusion

By following this comprehensive guide, you should be able to load your Yolov9 model trained on a custom dataset and make accurate predictions. Remember to double-check your dataset format, annotation files, and model loading to avoid common mistakes. If you still encounter issues, refer to the Yolov9 documentation and online resources for further assistance.

References

• Yolov9 Documentation: https://github.com/ultralytics/yolov9
• OpenCV Documentation: https://docs.opencv.org/4.5.5/
• Pascal VOC Annotation Format: https://wiki.eecs.yorku.ca/course_archive/2015-16/F/4415/_media/wiki:pascal-voc-annotation-format.pdf

Frequently Asked Question

Are you stuck with loading a YOLOv9 model trained on a custom dataset, only to encounter the frustrating “AttributeError: ‘str’ object has no attribute ‘shape'” error? Worry not, we’ve got you covered!