Per Detection class accuracy on validation set

[breannashi]

Search before asking

• [X] I have searched the YOLOv5 issues and discussions and found no similar questions.

Question

I have trained a yolov5 model, but I want to create a csv with tells me the label versus prediction class call for each detection with the name of its associated image also saved. detect --save-csv does not output all of this information. I have written a script that can generate this information from the label.txts but its inconvenient becuase I need to deal with all of the poor object detection edge cases when all I really care about it the class comparision for the boxes with the highest iou match. Please advise! Bree

Additional

nope

[github-actions[bot]]

👋 Hello @breannashi, thank you for your interest in YOLOv5 🚀! Please visit our ⭐️ Tutorials to get started, where you can find quickstart guides for simple tasks like Custom Data Training all the way to advanced concepts like Hyperparameter Evolution.

If this is a 🐛 Bug Report, please provide a minimum reproducible example to help us debug it.

If this is a custom training ❓ Question, please provide as much information as possible, including dataset image examples and training logs, and verify you are following our Tips for Best Training Results.

Requirements

Python>=3.8.0 with all requirements.txt installed including PyTorch>=1.8. To get started:

git clone https://github.com/ultralytics/yolov5  # clone
cd yolov5
pip install -r requirements.txt  # install

Environments

YOLOv5 may be run in any of the following up-to-date verified environments (with all dependencies including CUDA/CUDNN, Python and PyTorch preinstalled):

• Notebooks with free GPU:
• Google Cloud Deep Learning VM. See GCP Quickstart Guide
• Amazon Deep Learning AMI. See AWS Quickstart Guide
• Docker Image. See Docker Quickstart Guide

Status

If this badge is green, all YOLOv5 GitHub Actions Continuous Integration (CI) tests are currently passing. CI tests verify correct operation of YOLOv5 training, validation, inference, export and benchmarks on macOS, Windows, and Ubuntu every 24 hours and on every commit.

Introducing YOLOv8 🚀

We're excited to announce the launch of our latest state-of-the-art (SOTA) object detection model for 2023 - YOLOv8 🚀!

Designed to be fast, accurate, and easy to use, YOLOv8 is an ideal choice for a wide range of object detection, image segmentation and image classification tasks. With YOLOv8, you'll be able to quickly and accurately detect objects in real-time, streamline your workflows, and achieve new levels of accuracy in your projects.

Check out our YOLOv8 Docs for details and get started with:

pip install ultralytics

[glenn-jocher]

Hi Bree! 😊

It sounds like you're looking to refine the output of your detection results to focus on class comparison for high-confidence detections. You can achieve this by modifying the detect.py script to include additional logging for each detection that meets your criteria (e.g., highest IOU match).

Here’s a basic snippet you can add to your detect.py after detections are processed to filter and save the desired data:

import pandas as pd

# Assuming 'results' holds your model's output
detections = results.xyxy[0]  # detections for the first image in batch
high_conf_detections = detections[detections[:, 4] > confidence_threshold]  # confidence threshold, e.g., 0.25

# Prepare DataFrame
df = pd.DataFrame(high_conf_detections.numpy(), columns=['x1', 'y1', 'x2', 'y2', 'conf', 'class'])
df['image'] = 'name_of_image.jpg'  # Update with actual dynamic image name handling

# Save to CSV
df.to_csv('detections.csv', index=False)

This script snippet assumes you have a threshold for what you consider 'high confidence' and filters detections accordingly. You'll need to integrate it properly with the image handling in your existing detect.py script to dynamically assign image names.

For more detailed modifications, you might need to dive deeper into the post-processing steps of the detection script. If you need further guidance on this, feel free to check out the tutorials at https://docs.ultralytics.com/yolov5/tutorials/model_export/.

Hope this helps! 🚀

[breannashi]

This helps with editing the final prediction csv but I am still wondering how it will compare with the validation labels because when I use detection it seems to be agnostic of the fact that I have labels for those values.

[glenn-jocher]

@breannashi hi there!

To compare your detections with validation labels, you can modify the script to load the corresponding label files and perform a comparison. Here's a concise way to do it:

1. Load the label file for each image (assuming label files are in the same order as your images).
2. Compare the predicted classes and bounding boxes with the ground truth labels.
3. Calculate metrics like IOU to determine matches.

Here’s a basic example:

import numpy as np

# Load ground truth labels for an image
gt_labels = np.loadtxt('path_to_label_file.txt')

# Assuming 'detections' is your DataFrame from the previous example
for index, detection in detections.iterrows():
    # Compare detection['class'] with classes in gt_labels and calculate IOU, etc.
    # You can add conditions to filter matches based on IOU threshold

This approach requires you to align each detection with its corresponding label file and implement the logic for comparison. This might involve additional functions to calculate IOU or other metrics based on your specific needs.

Hope this points you in the right direction! 😊

[github-actions[bot]]

👋 Hello there! We wanted to give you a friendly reminder that this issue has not had any recent activity and may be closed soon, but don't worry - you can always reopen it if needed. If you still have any questions or concerns, please feel free to let us know how we can help.

For additional resources and information, please see the links below:

• Docs: https://docs.ultralytics.com
• HUB: https://hub.ultralytics.com
• Community: https://community.ultralytics.com

Feel free to inform us of any other issues you discover or feature requests that come to mind in the future. Pull Requests (PRs) are also always welcomed!

Thank you for your contributions to YOLO 🚀 and Vision AI ⭐