How to train objects365 without auto download the dataset.

[nocolour]

Dear all,

Have a nice day.

Since, I succeed to train on VisDrone. I wish to test on object365, but facing problem of downloading..

My Problem:
Object365 dataset too huge, download time out and can not complete with train.py --data Objects365.yaml.

To solve this problem, I use download manager to download the files manually one by one.

My question:

1. If I downloaded the object365 dataset manually. Where I need to unzip it? and folders structure?
2. If run python train.py --data Objects365.yaml, need to disable the download in Objects365.yaml ? If already downloaded manually.
3. How many free space needed for object365?
4. How many epochs needed? 300 enough?
5. Do I need to use --hpy hyp.finetune_objects365.yaml ? default='data/hyps/hyp.scratch.yaml'

Thanks for your help..

[glenn-jocher]

@nocolour I haven't had enough opportunity to train on Objects365 to answer your questions well, but yes naturally if you download it yourself you should comment out the download field in the yaml so it doesn't try to download it again. You should place your data in the structure indicated in the yaml here, and yes you can use either the default hyps or the 365 hyps you indicated to train, though you should not need 300 epochs since the dataset is much larger than COCO, which is where the 300 number comes from.

yolov5/data/Objects365.yaml

Lines 10 to 15 in fad57c2

	# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..]
	path: ../datasets/Objects365 # dataset root dir
	train: images/train # train images (relative to 'path') 1742289 images
	val: images/val # val images (relative to 'path') 5570 images
	test: # test images (optional)

[nocolour]

@nocolour I haven't had enough opportunity to train on Objects365 to answer your questions well, but yes naturally if you download it yourself you should comment out the download field in the yaml so it doesn't try to download it again. You should place your data in the structure indicated in the yaml here, and yes you can use either the default hyps or the 365 hyps you indicated to train, though you should not need 300 epochs since the dataset is much larger than COCO, which is where the 300 number comes from.

yolov5/data/Objects365.yaml

Lines 10 to 15 in fad57c2

	# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..]
	path: ../datasets/Objects365 # dataset root dir
	train: images/train # train images (relative to 'path') 1742289 images
	val: images/val # val images (relative to 'path') 5570 images
	test: # test images (optional)

Thank you for reply.

[nocolour]

@glenn-jocher
The download zip files, I don't see the label text files. Only have json file from zhiyuan_objv2_train.tar.gz.
zhiyuan_objv2_train.json is the label? Need to convert it? What I need to do?

[glenn-jocher]

@nocolour autodownload handles all conversion, I would recommend you simply use that:

python train.py --data Objects365.yaml

[nocolour]

@nocolour autodownload handles all conversion, I would recommend you simply use that:

python train.py --data Objects365.yaml

Ya, I found the way to solve it ready. But need do the test 1st. I will setup the localhost webserver to host the dataset I downloaded. Then change the download url in Objects365.yaml.

However still thank you.

[glenn-jocher]

@nocolour oh good idea!

[github-actions]

👋 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
• 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 ⭐!

[wangsun1996]

Can you provide a pre-training model on object365?

[glenn-jocher]

@wangsun1996 https://github.com/ultralytics/yolov5/releases/download/v6.0/yolov5m_Objects365.pt

[wangsun1996]

thank you very much! furthermore,Can you provide a pre-training model yolov5s6 or yolov5s on object365?

[glenn-jocher]

No, this is the only model available.

On Tue, 29 Nov 2022 at 18:59, wangsun1996 ***@***.***> wrote: thank you very much! furthermore,Can you provide a pre-training model yolov5s6 or yolov5s on object365? — Reply to this email directly, view it on GitHub <#4658 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AGMXEGKPVGW6PAS7G4S2XZDWK27IHANCNFSM5DLEYEIQ> . You are receiving this because you were mentioned.Message ID: ***@***.***>

-- <https://www.ultralytics.com/> *Glenn Jocher* Founder & CEO, Ultralytics +1 301 237 6695 <https://www.twitter.com/ultralytics> <https://www.youtube.com/ultralytics> <https://www.github.com/ultralytics> <https://www.linkedin.com/company/ultralytics>

[wangsun1996]

OK,thank you very much!!!

[wangsun1996]

i want to train a yolov5s6 on OBJ365 dataset to get a pre-train model，but we need 12h to train an epoch，Is there any way to speed up training?

[glenn-jocher]

👋 Hello! Thanks for asking about training speed issues. YOLOv5 🚀 can be trained on CPU (slowest), single-GPU, or multi-GPU (fastest). If you would like to increase your training speed some options are:

• Increase --batch-size
• Reduce --img-size
• Reduce model size, i.e. from YOLOv5x -> YOLOv5l -> YOLOv5m -> YOLOv5s
• Train with multi-GPU DDP at larger --batch-size
• Train on cached data: python train.py --cache (RAM caching) or --cache disk (disk caching)
• Train on faster GPUs, i.e.: P100 -> V100 -> A100
• Train on free GPU backends with up to 16GB of CUDA memory:

Good luck 🍀 and let us know if you have any other questions!