How to evaluate VisDrone-VID dataset?

[Wzh10032]

Search before asking

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

Question

How to evaluate Task 2: Object Detection in Videos in VisDrone-Dataset? The existing VisDrone.yaml seems to only support VisDrone-DET dataset.

Additional

No response

[UltralyticsAssistant]

👋 Hello @Wzh10032, thank you for your interest in YOLOv5 🚀! It looks like you are trying to evaluate the VisDrone-VID dataset. To assist you better, if this is a 🐛 Bug Report, please provide a minimum reproducible example to help us debug it.

If this is a custom training ❓ Question, it would be helpful to include additional details such as dataset image examples, any custom YAML files you might be using, and your training logs. Also, make sure you are following best practices outlined in our training guides.

Requirements

Ensure you have [Python>=3.8.0] with all necessary dependencies installed, including PyTorch>=1.8. You can start by cloning the YOLOv5 repository and installing the required packages:

git clone the yolov5 repository  # clone
cd into the yolov5 directory
pip install -r requirements.txt  # install

Environments

YOLOv5 can be run in various environments, including cloud-based notebooks with free GPU access, Google Cloud Deep Learning VMs, Amazon Deep Learning AMIs, and Docker images. This flexibility allows you to choose an environment that best suits your needs.

Status

Check the current status of YOLOv5 continuous integration tests in the GitHub Actions page to verify if all tests are passing.

This response is automated, but an Ultralytics engineer will be with you shortly to provide further assistance. If you have additional questions in the meantime, feel free to update this issue. 😊

[pderrenger]

@Wzh10032 thank you for your question! To evaluate Task 2: Object Detection in Videos (VisDrone-VID), you'll need to adapt the dataset configuration and ensure compatibility with the video evaluation process. Unfortunately, the current VisDrone.yaml is tailored for static image detection (VisDrone-DET). To proceed:

1. Prepare the video dataset in YOLO format by splitting video frames into images and ensuring proper annotations.
2. Create a custom YAML file to match your dataset's structure for videos.
3. For evaluation, use the same process as object detection but account for temporal aspects by evaluating predictions across frames.

For detailed guidance on dataset formatting, YOLO workflows, and evaluation, refer to the Ultralytics documentation. Let us know if you encounter specific issues!

[Wzh10032]

@Wzh10032 thank you for your question! To evaluate Task 2: Object Detection in Videos (VisDrone-VID), you'll need to adapt the dataset configuration and ensure compatibility with the video evaluation process. Unfortunately, the current VisDrone.yaml is tailored for static image detection (VisDrone-DET). To proceed:

1. Prepare the video dataset in YOLO format by splitting video frames into images and ensuring proper annotations.
2. Create a custom YAML file to match your dataset's structure for videos.
3. For evaluation, use the same process as object detection but account for temporal aspects by evaluating predictions across frames.

For detailed guidance on dataset formatting, YOLO workflows, and evaluation, refer to the Ultralytics documentation. Let us know if you encounter specific issues!

Thanks for your reply, I will try to solve this problem.

[pderrenger]

@Wzh10032 you're welcome! Let us know how it goes or if you face any challenges—happy to assist further. Best of luck with your VisDrone-VID evaluation!