Yolov8 results: AttributeError: 'list' object has no attribute 'path'- attributes not found

Question

I am working in colab with a Yolov8 object detection model. The code runs through the image subfolders correctly and the model is outputting correct results in the terminal, however I am unable to get results in a useable format after this point.

!pip install ultralytics
import ultralytics
from ultralytics import YOLO
import pandas as pd
import os

# Mount Google Drive to access local files
from google.colab import drive
drive.mount('/content/drive')

# Set model details
model = YOLO('/content/drive/MyDrive/train/weights/best.pt')

# Set path to the parent folder containing image subfolders
parent_folder = "/content/drive/MyDrive"

# Create an empty list to store the results
results_list = []

for image_file in image_files:
    # Set path to image file
    path_to_image = os.path.join(path_to_folder, image_file)

    # Predict using the model and get the bounding box coordinates
    results = model.predict(path_to_image, conf=0.8, classes=1, boxes=True,)
    
    # Extract information from  results
    image = results.path
    xyxy = results.xyxy
    conf = results.probs

    # Append the extracted information to the list
    results_list.append([image, xyxy, conf])
    
# Create a pandas DataFrame from the results list
df = pd.DataFrame(results_list, columns=['image', 'xyxy', 'conf'])

# Save the DataFrame as a CSV file
df.to_csv('/content/drive/MyDrive/1.predictions/object_detection_results.csv', index=False)

I have tried getting things into a json, txt or csv format but can't get extract any of the variables needed, mainly xyxy.

For the use case, it is important to get:

1. the image name
2. bounding box dimensions
3. confidence interval

ideally as separate columns, and one row per prediction.

I am getting the following error:

AttributeError                            Traceback (most recent call last)
<ipython-input-26-bac2882f463a> in <cell line: 20>()
     26 
     27     # Extract information from  results
---> 28     imagename = results.path
     29     boxes = results.boxes
     30     probs = results.probs

AttributeError: 'list' object has no attribute 'path'

The same error occurs for boxes / probs and xyxy.

However I checked the documentation and 'path' / 'boxes' and 'probs' seem to be the correct attributes.

At a loss here, any advice to get a semi useable format would be appreciated.

Answer 1

The results here is a list of ultralytics.engine.results.Results class objects, a class for storing and manipulating inference results. Each object in this list represents result information for every image in a source. As you pass to the model a single image at a time, you can refer to the [0] index of this list to get all the needed information.

xyxy and conf are boxes properties, they are both stored as torch.Tensor, but you can translate them to lists for convenience.

image = results[0].path
xyxy = results[0].boxes.xyxy.tolist()
conf = results[0].boxes.conf.tolist()

More information about working with results here.