Documentation

# FAQ - Frequently Asked Questions

## Technical Issues

### Why can't I access the internet when I am running the ai-vision-solution-software?
**One possible reason is, that your device and one of the docker instances/containers have the same local ip address.**

To resolve this, start by checking the instances' IP addresses:
![docker inspect command to find out the container addresses](https://imaginghub.com/uploads/editor/images/2020/7aed2528-bfeb-4a69-b4a5-37afc5c6759e.jpeg)

If one of the IP addresses shown matches your Solution Kit's external IP address, you can run into connection problems. However, it is always a good idea to configure the docker daemon to use an internal IP address pool, which doesn't match your company's subnets. See [this docker community forum entry][1] for more details.

###I feel like my Jetson Nano is running too hot
NVidia recommends to use a fan when the board is operated under unfavorable conditions (no ventilation, hot environment) under high load.

To check if you need a fan, run the **tegrastats** command on the nano.

If the listed temperatures reach the limits specified in the [official hardware documentation ][2], you can attach a 5V fan to the fan header for improved 
performance.

[This one][3] is recommended by NVidia.

###The WebGUI does not change models properly or does not show an image

**This is a temporary workaround and will be fixed in a future release!**

In some cases the image processing container crashes after running for a long period of time.

List the running docker containers by running

docker ps
from the Jetson's terminal.

Check if the image-processing container is running.
If it is not present, you have two options to workaround that problem.

1. Have a look at the [Solution Kit Download area](https://imaginghub.com/projects/438-ai-vision-solution-kit-with-aws-cloud-connection#files-section) to find and flash a current (20200908 or later) software image to your Jetson Nano. See also the "*I totally bricked my Solution Kit!*" FAQ entry.
 2. Attach a USB-device of your choice to one of the Nano's USB ports and restart the AI-Vision-Solution Kit using the **ai-vision-solution-kit restart** command. The image-processing container will run stable with a USB-device attached.

### I totally bricked my Solution Kit!
That is not a big deal. You have two options here.

1. Contact your local Basler sales person or the Basler product support.
 2. Visit [our Solution Kit download area](https://imaginghub.com/projects/438-ai-vision-solution-kit-with-aws-cloud-connection#files-section). Here you can find a file with the title *Basler AI Vision Solution Kit Image*. This is the current software image we ship with newly ordered Solution Kits. You can use this image to flash your Solution Kit.

If you need further details on how to flash the Solution Kit, please refer to NVIDIA's manuals on [Writing Image to the microSD Card][4].

## Usability

### Locally or remotely - what is the intended way to connect to the AI Solution Kit?
Local usage - i.e. directly connecting via HDMI/keyboard/mouse - decreases the amount of computing power and memory available to your application, significantly. So we highly recommend to use it as a remote kit. That implies connecting to the Kit via SSH (for console access) and our provided Web GUI.

###I want to be up to date. What do I have to do?
Visit [our Solution Kit download area](https://imaginghub.com/projects/438-ai-vision-solution-kit-with-aws-cloud-connection#files-section) regularly to check for new Solution Kit software packages. Please install them according to the instructions in Step 3 of our [Starter Documentation](https://imaginghub.com/projects/440-starter-documentation).

Generally, the docker containers will be updated automatically when starting the system. However, this only works with the following prerequisites:

1. The board is connected to the internet and is able to connect to web services on port 80 and 443.
 2. The system time is set correctly. Usually the board will automatically sync the time via NTP. Please check the correct UTC date/time by calling `date` on the board.

If the system time cannot be set automatically due to network restrictions, please consider the following options:

- Set up the board to sync with your company internal time server (if existing)
 - Set the time manually by calling `sudo date --set`, for example: `sudo date --set="23 Dec 2020 10:12:34"` . after setting the correct time, please execute `ai-vision-solution kit restart` to update to the latest released versions of the docker images.

## Other Issues

### What types of objects does the Office Object Detection recognize and how can I extend it?
The Office Object Detection contains the following types of objects:

- computer keyboard
 - computer mouse
 - chair
 - computer monitor
 - desk
 - laptop

If you plan to extend the model or to change the types of detected objects by retraining it, please refer to our corresponding documentation on [How to create, train and deploy a Machine Learning Model](https://imaginghub.com/projects/441-how-to-create-train-and-deploy-a-machine-learning-model) or have a look at [the sample code](https://github.com/basler/ai-vision-solution-kit-samples/tree/master/sagemaker_neo_compilation_jobs_jetson_nano/train_object_detection_models/ssd_officeitems_detection_gluoncv).

[1]: https://forums.docker.com/t/custom-ip-range-for-new-networks/60839
  [2]: https://docs.nvidia.com/jetson/l4t/index.html#page/Tegra%2520Linux%2520Driver%2520Package%2520Development%2520Guide%2Fpower_management_nano.html%23wwpID0E03E0HA
  [3]: https://noctua.at/en/nf-a4x20-5v-pwm
  [4]: https://developer.nvidia.com/embedded/learn/get-started-jetson-nano-devkit#write

FAQ - Frequently Asked Questions

Technical Issues

Why can't I access the internet when I am running the ai-vision-solution-software?

One possible reason is, that your device and one of the docker instances/containers have the same local ip address.

To resolve this, start by checking the instances' IP addresses: docker inspect command to find out the container addresses

If one of the IP addresses shown matches your Solution Kit's external IP address, you can run into connection problems. However, it is always a good idea to configure the docker daemon to use an internal IP address pool, which doesn't match your company's subnets. See this docker community forum entry for more details.

I feel like my Jetson Nano is running too hot

NVidia recommends to use a fan when the board is operated under unfavorable conditions (no ventilation, hot environment) under high load.

To check if you need a fan, run the tegrastats command on the nano.

If the listed temperatures reach the limits specified in the official hardware documentation , you can attach a 5V fan to the fan header for improved performance.

This one is recommended by NVidia.

The WebGUI does not change models properly or does not show an image

This is a temporary workaround and will be fixed in a future release!

In some cases the image processing container crashes after running for a long period of time.

List the running docker containers by running

docker ps

from the Jetson's terminal.

Check if the image-processing container is running. If it is not present, you have two options to workaround that problem.

Have a look at the Solution Kit Download area to find and flash a current (20200908 or later) software image to your Jetson Nano. See also the "I totally bricked my Solution Kit!" FAQ entry.
Attach a USB-device of your choice to one of the Nano's USB ports and restart the AI-Vision-Solution Kit using the ai-vision-solution-kit restart command. The image-processing container will run stable with a USB-device attached.

I totally bricked my Solution Kit!

That is not a big deal. You have two options here.

Contact your local Basler sales person or the Basler product support.
Visit our Solution Kit download area. Here you can find a file with the title Basler AI Vision Solution Kit Image. This is the current software image we ship with newly ordered Solution Kits. You can use this image to flash your Solution Kit.

If you need further details on how to flash the Solution Kit, please refer to NVIDIA's manuals on Writing Image to the microSD Card.

Usability

Locally or remotely - what is the intended way to connect to the AI Solution Kit?

Local usage - i.e. directly connecting via HDMI/keyboard/mouse - decreases the amount of computing power and memory available to your application, significantly. So we highly recommend to use it as a remote kit. That implies connecting to the Kit via SSH (for console access) and our provided Web GUI.

I want to be up to date. What do I have to do?

Visit our Solution Kit download area regularly to check for new Solution Kit software packages. Please install them according to the instructions in Step 3 of our Starter Documentation.

Generally, the docker containers will be updated automatically when starting the system. However, this only works with the following prerequisites:

The board is connected to the internet and is able to connect to web services on port 80 and 443.
The system time is set correctly. Usually the board will automatically sync the time via NTP. Please check the correct UTC date/time by calling date on the board.

If the system time cannot be set automatically due to network restrictions, please consider the following options:

Set up the board to sync with your company internal time server (if existing)
Set the time manually by calling sudo date --set, for example: sudo date --set="23 Dec 2020 10:12:34" . after setting the correct time, please execute ai-vision-solution kit restart to update to the latest released versions of the docker images.

Other Issues

What types of objects does the Office Object Detection recognize and how can I extend it?

The Office Object Detection contains the following types of objects:

computer keyboard
computer mouse
chair
computer monitor
desk
laptop

If you plan to extend the model or to change the types of detected objects by retraining it, please refer to our corresponding documentation on How to create, train and deploy a Machine Learning Model or have a look at the sample code.

Info