urllib.error.URLError in downloading the sample dataset

[May Phyu Khin]

I am starting user of NNC.
I downloaded NNC for window app, authenticated my account, and tried to download the sample data set, but, urllib.error.URLError occurs.
Please help me how to fix this error.
Thank you all for your kindness and support.

[Tomonobu Tsujikawa]

Hi,

What error messages do you get?

I just tried download mnist on my environment, and it works with NNC v3.1.0.

(You can skip this if you don't use proxy)

If you cannot connect servers, you may need to set the proxy information by HTTP_PROXY and HTTPS_PROXY environment variables to connect Internet.

You can set these information at the Engine tab in Setup menu at the top right corner.

The download scripts are available here:

{install_path}/neural_network_console/libs/plugins/_Pre_Process/_Download_or_Generate_Dataset/

create_mnist_csv.py will download mnist sample dataset.

Please try using your browser to see if you can download from these URLs in this script.

2024年4月16日火曜日 9:55:23 UTC+9 mayphyuk...@gmail.com:

[May Phyu Khin]

Dear Tomonobu Tsujikawa 様

Thank you very much for your reply.

I solved my problem by using the solutions from the following link.

https://groups.google.com/g/neural_network_console_users_jp/c/mTnWKYtEgc0

May I also know for the image detection, is around 150 training images will suitable for demo version of image detection?

May I also know is it possible for multiple objects detection within one frame by using NNC?

Respectfully,

May Phyu Khin

[Tomonobu Tsujikawa]

Hi,

> May I also know for the image detection, is around 150 training images will suitable for demo version of image detection?

It's depends on complexity, but I feel 150 images are too small to train from the scratch.

Even in image binary classification tutorial, the training uses 1,500 images.

So, please check learning curve and add more data if needed.

> May I also know is it possible for multiple objects detection within one frame by using NNC?

Here is a list of sample projects for the Neural Network Console (Cloud):

https://dl.sony.com/project/

I think "Object detection using CenterNet. Backbone is ResNet-18." is multiple object detection.

This sample project is also available on Neural Network Console (Windows).

This training dataset contains 90,000 images.

2024年4月18日木曜日 9:47:06 UTC+9 mayphyuk:

Dear Tomonobu Tsujikawa 様

Thank you very much for your reply.

I solved my problem by using the solutions from the following link.

https://groups.google.com/g/neural_network_console_users_jp/c/mTnWKYtEgc0

May I also know for the image detection, is around 150 training images will suitable for demo version of image detection?

May I also know is it possible for multiple objects detection within one frame by using NNC?

Respectfully,

May Phyu Khin

On Tuesday, April 16, 2024 at 11:59:55 AM UTC+9 tomonobu wrote:

Hi,

What error messages do you get?

I just tried download mnist on my environment, and it works with NNC v3.1.0.

(You can skip this if you don't use proxy)

If you cannot connect servers, you may need to set the proxy information by HTTP_PROXY and HTTPS_PROXY environment variables to connect Internet.

You can set these information at the Engine tab in Setup menu at the top right corner.

The download scripts are available here:

{install_path}/neural_network_console/libs/plugins/_Pre_Process/_Download_or_Generate_Dataset/

create_mnist_csv.py will download mnist sample dataset.

Please try using your browser to see if you can download from these URLs in this script.

2024年4月16日火曜日 9:55:23 UTC+9 mayphyuk:

[May Phyu Khin]

Dear Tomonobu Tsujikawa 様

Thank you very much for your reply.

May I also know that for the semantic detection, can we use the transfer learning at NNC?

If so, may I know the tutorial link.

Respectfully,

May Phyu Khin