Do you have any thoughts about reforming YOLOV5?

jamjamjon commented 3 years ago

Search before asking

[X] I have searched the YOLOv5 issues and found no similar feature requests.

Description

Do you have an idea to make a big modify with yolov5? Like changing backbone, detect head, even removing the anchors like YOLOX. YOLOV5 is powerful right now, But I just want to see some big change in a short future. If you have any thoughts about it, please let me know. I really like your work in YOLOV5! Have a good day!

Use case

No response

Additional

No response

Are you willing to submit a PR?

[ ] Yes I'd like to help by submitting a PR!

github-actions[bot] commented 3 years ago

👋 Hello @Jamjamjamjon, 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 screenshots and minimum viable code to reproduce your issue, otherwise we can not help you.

If this is a custom training ❓ Question, please provide as much information as possible, including dataset images, training logs, screenshots, and a public link to online W&B logging if available.

For business inquiries or professional support requests please visit https://ultralytics.com or email Glenn Jocher at glenn.jocher@ultralytics.com.

Requirements

Python>=3.6.0 with all requirements.txt installed including PyTorch>=1.7. To get started:

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

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):

Google Colab and Kaggle 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 (train.py), validation (val.py), inference (detect.py) and export (export.py) on MacOS, Windows, and Ubuntu every 24 hours and on every commit.

zhiqwang commented 3 years ago

IMO, The paradigm currently adopted by YOLOv5 is very refined, and this approach is also highly scalable and configurable. On the other hand, you may also see some refactored versions of yolov5 in the GitHub community with the backbone, fpn(PAN) and head forms, such as the following

EDIT: I think I've misunderstood this ticket, just ignore my comment above :/

glenn-jocher commented 2 years ago

@Jamjamjamjon yes on the architecture side there is of course a lot of experimentation we could do, but a lot of this is high risk basic research that may not produce improvements.

BUT, there is a lot of low risk development also we need to do, mostly related to MLops/Devops capabilities that should produce benefits for everyone:

Single and Multi-GPU CI
Major code refactor to create 1) multi-backend (i.e. PyTorch, TF, Keras, TFLite, CoreML, torchscript) model loading, 2) multi-backend inference functions. This would allow detect.py and val.py to share model loading and inference functions, and then this would allow val.py to test exported models (TFLite, ONNX, CoreML etc.), which multiple users have been asking for. This would really help users verify that their deployed models are performing well, they could quantify drops in performance from quantization etc. which we don't offer now.
Better introspection tools in val.py to highlight problems, i.e. edge cases, poorly performing images, poorly performing classes. This would help inform dataset updates to target weak points. For example in COCO some classes perform very poorly, and I'm sure some images in particular perform poorly, we want to be able to display those problem images, which we can't do now since metrics are not computed per image.

EDIT: Another big point is we'd like to expand the tasks we target to classification and segmentation. With classification we've started a branch at https://github.com/ultralytics/yolov5/blob/classifier/classifier.py, with segmentation we still have a lot more ways to go (new segmentation head, loss function, plots/annotation updates).

jamjamjon commented 2 years ago

thanks！I will make some change with model architecture and do some experiments. I will let you know once I get something. Thanks again！Have a good day！

发自我的iPhone

------------------ Original ------------------ From: Glenn Jocher @.> Date: Thu, Nov 4, 2021 8:05 PM To: ultralytics/yolov5 @.> Cc: Jamjamjamjon @.>, Mention @.> Subject: Re: [ultralytics/yolov5] Do you have any thoughts about reformingYOLOV5? (Issue #5495)

@Jamjamjamjon yes on the architecture side there is of course a lot of experimentation we could do, but a lot of this is high risk basic research that may not produce improvements.

BUT, there is a lot of low risk development also we need to do, mostly related to MLops/Devops capabilities that should produce benefits for everyone:

Single and Multi-GPU CI

Major code refactor to create 1) multi-backend model loading, 2) multi-backend inference functions. This would allow detect.py and val.py to share model loading and inference functions, and then this would allow val.py to test exported models (TFLite, ONNX, CoreML etc.), which multiple users have been asking for.

Better introspection tools in val.py to highlight problems, i.e. edge cases, poorly performing images, poorly performing classes. This would help inform dataset updates to target weak points. For example in COCO some classes perform very poorly, and I'm sure some images in particular perform poorly, we want to be able to display those problem images, which we can't do now since metrics are not computed per image.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android.

glenn-jocher commented 2 years ago

@Jamjamjamjon great! Yes please help future uses by submitting PRs for any improvements or bug fixes you come across.

github-actions[bot] commented 2 years ago

👋 Hello, this issue has been automatically marked as stale because it has not had recent activity. Please note it will be closed if no further activity occurs.

Access additional YOLOv5 🚀 resources:

Wiki – https://github.com/ultralytics/yolov5/wiki
Tutorials – https://docs.ultralytics.com/yolov5
Docs – https://docs.ultralytics.com

Access additional Ultralytics ⚡ resources:

Ultralytics HUB – https://ultralytics.com/hub
Vision API – https://ultralytics.com/yolov5
About Us – https://ultralytics.com/about
Join Our Team – https://ultralytics.com/work
Contact Us – https://ultralytics.com/contact

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 YOLOv5 🚀 and Vision AI ⭐!

ultralytics / yolov5