Yolov5: Can not download pre-trained weights in China.

[shihaoyin]

The Google Drive can not access from China, maybe a BaiduYun link is good to Chinese people.

[github-actions[bot]]

Hello @shihaoyin, thank you for your interest in our work! Please visit our Custom Training Tutorial to get started, and see our Jupyter Notebook , Docker Image, and Google Cloud Quickstart Guide for example environments.

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 model or data training question, please note that Ultralytics does not provide free personal support. As a leader in vision ML and AI, we do offer professional consulting, from simple expert advice up to delivery of fully customized, end-to-end production solutions for our clients, such as:

• Cloud-based AI systems operating on hundreds of HD video streams in realtime.
• Edge AI integrated into custom iOS and Android apps for realtime 30 FPS video inference.
• Custom data training, hyperparameter evolution, and model exportation to any destination.

For more information please visit https://www.ultralytics.com.

[glenn-jocher]

@shihaoyin thank you for the bug report! This is a big problem indeed. We used to host weights on a GCP bucket which was able to be used in China, but Google was charging us monthly per download, so we switched to the current solution.

Perhaps we can update attempt_download() function to try Baidu if Google Drive is not reachable.

Are there any volunteers who might be able to do this and submit a PR?

https://github.com/ultralytics/yolov5/blob/c6fc4242a21cac23526c221994eff0886e303e0e/utils/google_utils.py#L10-L32

[HenryWang628]

The Google Drive can not access from China, maybe a BaiduYun link is good to Chinese people.
@shihaoyin
you can try this
链接：https://pan.baidu.com/s/1Zk2Ksfl_v-apbRBQ_mqc6w
提取码：00mp

[shihaoyin]

The Google Drive can not access from China, maybe a BaiduYun link is good to Chinese people.
@shihaoyin
you can try this
链接：https://pan.baidu.com/s/1Zk2Ksfl_v-apbRBQ_mqc6w
提取码：00mp

Download successfully，thanks a lot。

[glenn-jocher]

Are AWS (Amazon) S3 buckets/files accessible from China?

[shihaoyin]

Are AWS (Amazon) S3 buckets/files accessible from China?

Not yet tested, maybe you can paste a file link on AWS S3, and I will test it.

[glenn-jocher]

@shihaoyin is this still a problem? I updated the code to try to fix this problem, so I believe the auto-download functionality should work for the entire world now, including China.

[Lornatang]

@glenn-jocher I am in Shanghai,China, the option is False. Wish us good luck!

[Samuel-wei]

The Google Drive can not access from China, maybe a BaiduYun link is good to Chinese people.
@shihaoyin
you can try this
链接：https://pan.baidu.com/s/1Zk2Ksfl_v-apbRBQ_mqc6w
提取码：00mp

I get it, thanks.

[glenn-jocher]

@Lornatang can you try to download these weights from Shanghai? I'm experimenting with a new weight hosting method :)

import torch

# method 1
torch.hub.download_url_to_file('https://storage.googleapis.com/ultralytics/yolov5/ckpt/yolov5x.pt', 'yolov5x_1.pt')

# method 2
torch.hub.download_url_to_file('https://github.com/ultralytics/yolov5/releases/download/v2.0/yolov5x.pt', 'yolov5x_2.pt')

[Lornatang]

@glenn-jocher Yes!

(base) ➜  ~ wget https://storage.googleapis.com/ultralytics/yolov5/ckpt/yolov5x.pt
--2020-08-01 10:44:46--  https://storage.googleapis.com/ultralytics/yolov5/ckpt/yolov5x.pt
正在解析主机 storage.googleapis.com (storage.googleapis.com)... 34.64.4.112, 34.64.4.80, 34.64.4.16
正在连接 storage.googleapis.com (storage.googleapis.com)|34.64.4.112|:443... 已连接。
已发出 HTTP 请求，正在等待回应... 200 OK
长度： 178507455 (170M) [application/octet-stream]
正在保存至: “yolov5x.pt”

yolov5x.pt            1%[                    ]   2.82M  1.31MB/s               ^Z
[1]  + 21731 suspended  wget https://storage.googleapis.com/ultralytics/yolov5/ckpt/yolov5x.pt

[glenn-jocher]

Great! What about method 2 (different url):

import torch

# method 2
torch.hub.download_url_to_file('https://github.com/ultralytics/yolov5/releases/download/v2.0/yolov5x.pt', 'yolov5x_2.pt')

[Lornatang]

@glenn-jocher Yes. This is the method I have been using in projects

 (base) ➜  ~ wget https://github.com/ultralytics/yolov5/releases/download/v2.0/yolov5x.pt
--2020-08-01 12:49:11--  https://github.com/ultralytics/yolov5/releases/download/v2.0/yolov5x.pt
正在解析主机 github.com (github.com)... 140.82.113.3, 192.30.253.112
正在连接 github.com (github.com)|140.82.113.3|:443... 已连接。
已发出 HTTP 请求，正在等待回应... 302 Found
位置：https://github-production-release-asset-2e65be.s3.amazonaws.com/264818686/99d9dd80-d2bf-11ea-8ff8-a4f0684ba6c2?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Credential=AKIAIWNJYAX4CSVEH53A%2F20200801%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20200801T044915Z&X-Amz-Expires=300&X-Amz-Signature=73d2e672d61d51d53fb65837b66f76cb441d876a4f07a4641301eb9f31d7b45c&X-Amz-SignedHeaders=host&actor_id=0&repo_id=264818686&response-content-disposition=attachment%3B%20filename%3Dyolov5x.pt&response-content-type=application%2Foctet-stream [跟随至新的 URL]
--2020-08-01 12:49:15--  https://github-production-release-asset-2e65be.s3.amazonaws.com/264818686/99d9dd80-d2bf-11ea-8ff8-a4f0684ba6c2?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Credential=AKIAIWNJYAX4CSVEH53A%2F20200801%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20200801T044915Z&X-Amz-Expires=300&X-Amz-Signature=73d2e672d61d51d53fb65837b66f76cb441d876a4f07a4641301eb9f31d7b45c&X-Amz-SignedHeaders=host&actor_id=0&repo_id=264818686&response-content-disposition=attachment%3B%20filename%3Dyolov5x.pt&response-content-type=application%2Foctet-stream
正在解析主机 github-production-release-asset-2e65be.s3.amazonaws.com (github-production-release-asset-2e65be.s3.amazonaws.com)... 52.216.229.219
正在连接 github-production-release-asset-2e65be.s3.amazonaws.com (github-production-release-asset-2e65be.s3.amazonaws.com)|52.216.229.219|:443... 已连接。
已发出 HTTP 请求，正在等待回应... 200 OK
长度： 178507455 (170M) [application/octet-stream]
正在保存至: “yolov5x.pt.1”

yolov5x.pt.1          0%[                    ]  24.58K   940 B/s    剩余 2d 4h ^C

For speed, I recommend you to use the first method.

[glenn-jocher]

Ok great! What speed difference do you see between the two methods?

[Lornatang]

Take users in Mainland China as an example:

If you are not using a VPN, use

import torch

# method 1
torch.hub.download_url_to_file('https://storage.googleapis.com/ultralytics/yolov5/ckpt/yolov5x.pt', 'yolov5x_1.pt')

else

import torch

# method 2
torch.hub.download_url_to_file('https://github.com/ultralytics/yolov5/releases/download/v2.0/yolov5x.pt', 'yolov5x_2.pt')

Both methods can get the same weight, it's just a matter of speed.

[Lornatang]

@glenn-jocher
I see a huge difference, it may be the difference between 5M and 50K.

[glenn-jocher]

@Lornatang ok, thanks!