Installation of HSA KIT running on a single machine, fully confined within the internal network.

For further installation options, see HSA KIT Deployment

Minimum Hardware Requirements

Annotation Only

  • 8th Gen Intel i5 CPU / AMD Ryzen 5 3600
  • 16 GB RAM
  • 10 GB SSD harddrive space for installation
  • Enough free space on harddrive to store any images to analyse OR
  • 100 mBit connection to NAS where files are stored, a gigabit connection is strongly recommended.
  • Windows 10 or later

AI Training

  • CUDA 10 capable NVIDIA GPU with a minimum of 6 GB Dedicated gRAM

Software Requirements

  • Chrome or Edge Browser installed
  • Java SDK and C++ Redistributables will be installed automatically during HSA KIT installation

Basic Installation

  1. Download the HSA KIT software. This is a large file, so it is provided as a .zip archive. Unzip the downloaded file to access its contents.
  2. Run the setup file with administrator rights. It is recommended to use the default settings during the installation process.
  3. After the installation is complete, restart your Windows system to ensure all changes take effect.
  4. Once your system has restarted, locate the shortcut for HSA KIT that was created during the installation. Launch the application using this shortcut.
  5. When HSA KIT launches, it should automatically open the Chrome browser. If a different browser opens instead, set Chrome as the default browser in your system settings.

Which version should I choose?

HSA KIT comes in two versions: Cuda 10 and Cuda 11, with each version being intened for different hardware.

As a rule of thumb, any NVIDIA graphics card beignning with GTX will need CUDA 10 any on beginning with RTX will need CUDA 11.

To determine whether your GPU supports CUDA 10 or CUDA 11, you can follow these steps:

  1. Identify your GPU model: Find out the specific model of your GPU. This information can typically be obtained through the manufacturer’s documentation, the GPU’s packaging, or by checking your computer’s specifications if the GPU is pre-installed.
  2. Visit the GPU vendor’s website
  3. Find the GPU’s „compute capability“
  4. A compute capablilty of 7.5 or higher will allow the use of CUDA 11

If you are having trouble finding the necessary information on the vendor’s website, you can also consult HSA, the GPU’s user manual, or reach out to the manufacturer’s support team for assistance.

Initial HSA KIT Setup

  1. Locate the shortcut for HSA KIT that was created during the installation. Launch the application using this shortcut.
  2. When HSA KIT launches, it should automatically open the Chrome browser. If a different browser opens instead, set Chrome as the default browser in your system settings.
  3. You will be transferred to the licensing page
  4. Send us the Hardware ID so that we can generate the License Key. The hardware ID can be found at the top of the licensing page and looks similar to this: XXXXXX-XXXXXXX-XXXXXXX-XXXXXXX
  1. Once you have received the License Key from us, enter it.
  2. Configure the HSA KIT folders to your liking:
    • Storage Path
      • The location where your images you want to analyze are.
      • Should have enough space to accomaodate all files to analyse.
      • Do not use a loaction with many irrelevant files.
      • For a shared computer, this location must be accessible for all users.
    • Data Path
      • The location where most results and projects are stored.
      • To prevent loss of work, include this location in your backups.
      • For a shared computer, this location must be accessible for all users.
    • Gallery Path
      • The location where gallery output is saved.
      • Allows to export many smaller images of individual annotations.
      • For a shared computer, this location must be accessible for all users.
    • Java Path
      • The location of your java sdk installation.
      • Usually, the default value is fine.
  3. Click on „SUBMIT“ to confirm You should then be able to log in.

Can I use HSA KIT offline?

Yes.

Normal use of HSA KIT with an internet connection will not require the next steps.

To use the offline pretrained models, please follow the installation instructions below:

  1. Download the collected pretrained models from here.
  2. Once the download is complete, locate the downloaded „.cache.zip“ file on your computer.
  3. Unpack the zip folder in your own user directory. For example, if your username is „UserName,“ extract the contents of the „.cache.zip“ file to the following path: „C:\Users\UserName“. You should now have a folder named „.cache“ in your user directory.
  4. Verify that the folder „C:\Users\UserName\.cache\torch\hub\checkpoints“ exists and contains several checkpoints. These checkpoints are the pretrained models you will be using.
  5. Make adjustments if necessary: Ensure that the folder structure is correct and matches the following path: „C:\Users\UserName\.cache\torch\hub\checkpoints“. If you encounter any issues or errors, double-check that the „.cache.zip“ file was unpacked correctly, and the checkpoints are in the correct directory.

Common errors and their solutions:

  • Error: „.cache.zip“ not unzipped. Solution: Unpack the „.cache.zip“ file following steps 2 and 3 of the installation instructions. Make sure to extract the contents to the correct user directory.
  • Error: „.cache.zip“ unpacked in „.cache“ folder, resulting in a nested folder structure (e.g., „C:\Users\UserName\.cache\.cache\torch\hub\checkpoints“). Solution: Ensure that the „.cache.zip“ file is correctly unzipped, and the resulting folder structure matches step 3 of the installation instructions. The correct path should be „C:\Users\UserName\.cache\torch\hub\checkpoints“.

By following these installation instructions and addressing common errors, you should be able to successfully set up and use the offline pretrained models.