Same yolov5s training, but one over-fitting and one training is very good.

[mimimind]

Search before asking

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

Question

When I was using the yolov5s model, when I used the same data set (rdd2022) for road crack detection training, the same yolov5s, the same hyperparameter, the same graphics card (A4000-16g), but at the beginning of this month, I trained very well. In the past two days, I have been training and found that I have been over-fitting.I looked for a lot of ways but didn't find the problem. I don't know if anyone can help me solve the problem.

Additional

No response

[github-actions]

👋 Hello @mimimind, 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]

@mimimind hello there! 😊 It sounds a bit puzzling that you're experiencing different training outcomes with the same setup. A few things to check that might explain the variance:

1. Data Split: Ensure the training-validation split is consistent between runs. Inadvertent changes can cause discrepancies in training results.
2. Environment: Verify that the software environment (PyTorch version, CUDA version, etc.) hasn't changed. Differences here can affect model performance.
3. Random Seeds: YOLOv5 training involves randomness (e.g., data shuffling). Setting a fixed seed can help ensure consistency across training sessions.
4. Updates in YOLOv5 Repo: Even if your setup hasn't changed, updates to the YOLOv5 repository might have occurred. Ensure you're training with the same commit/version of YOLOv5 for both runs.

Here's a tiny bit of code to fix the seed, just in case:

import torch
torch.manual_seed(42)  # Use a consistent seed

If after checking these factors the issue still persists, the detailed logs of both training runs might offer some clues. Comparing them might reveal subtle differences not immediately apparent.

Hope this helps point you in the right direction!