Getting Started

First launch, license activation, runtime setup, and creating your first project

Installation

Step 1: Installation using the Csmart Installer

• Click on the following link and download the csmart-studio installer for Windows.

• Locate the downloaded installer file and double-click it to begin the installation.

• Ignore any Windows or antivirus messages and proceed with the installation.

Important: Select the "Only Me" option during installation to avoid requiring admin privileges for software use.

• Once installation is complete, a Csmart Studio desktop shortcut will be created.

First Launch

When you open Studio Desktop for the first time, the application checks for a valid license. If no license is found, you are redirected to the License Management screen automatically.

Step 2: Activate Your License

1. On the License Management screen, click Select License File.

2. In the file dialog, navigate to your .lic license file and select it.

3. The application verifies the license signature and expiration date.

4. If valid, the status changes to Valid and the full application becomes available.

The license file is copied to the application's internal data folder. You only need to perform this step once per installation. The license persists across application updates.

If your license has expired or is invalid, all screens except License Management will be disabled. Contact Csmart to obtain a renewed license file.

Step 3: Install the Python Runtime

Studio Desktop requires a Python runtime to execute training, testing, and export operations. The runtime is a self-contained Python distribution — it does not interfere with any existing Python installation on your system.

1. Navigate to Hardware Settings from the sidebar.

2. The application checks whether the required Python version is installed.

3. If the runtime is missing or outdated, click Install Runtime.

4. The download and extraction progress is displayed in real time.

5. Once complete, the status updates to Installed with the version number.

The runtime is installed to the application's user data folder and is approximately 1-2 GB in size. An internet connection is required for the initial download.

Step 4: Verify GPU Availability

If you have an NVIDIA GPU installed, you can verify that it is correctly detected:

1. On the Hardware Settings screen, click Check GPU.

2. The application spawns a quick script that probes CUDA availability.

3. The result shows your GPU name and CUDA version, or indicates that no GPU was found.

Training without a GPU is possible but significantly slower. If your GPU is not detected, ensure that NVIDIA drivers and the CUDA toolkit are properly installed. Refer to the GPU Installation guide for details.

The Home Screen

After license activation and runtime setup, the Home screen is your starting point. It displays six action cards:

Card	Description
New Project	Create a fresh project with a new dataset configuration
Browse Projects	Open, edit, or delete existing projects
Split Dataset	Jump directly to dataset splitting (requires an open project)
Train AI Model	Jump to training (requires an open project)
Test AI Model	Jump to testing (requires an open project)
Export AI Model	Jump to export (requires an open project)

The bottom of the screen shows two counters: Models Exported (total exports across all projects) and Training Hours (cumulative time spent training).

Cards that require an open project are disabled until you create or open one. Start by clicking New Project or Browse Projects.

Creating a New Project

1. From the Home screen, click New Project.

2. Fill in the project details:

Project Name

Enter a clear, descriptive name for the project. The name is used as the folder name on disk.

• Use lowercase letters, numbers, and hyphens only.

• Minimum 3 characters.

• Do not include version numbers — versioning is handled automatically by the application.

Examples: arabica-brazil-2026, robusta-defect-classifier, sca-green-grading

Coffee Metadata (Optional)

You can attach metadata to the project to keep track of the coffee being classified. All fields are optional:

Field	Description	Example Values
Species	Coffee species (dropdown)	Arabica, Robusta (Canephora), Liberica, Excelsa
Variety	Coffee variety	Bourbon, Typica, Mixed
Origin	Country of origin	Brazil, Colombia, Ethiopia
Region	Growing region	Sul de Minas, Huila, Yirgacheffe
Processing	Processing method	Natural, Washed, Honey

This metadata is stored in project.yaml and can be edited later from the Browse Projects screen.

Project Location

Click Select Folder to choose where the project will be created on your computer.

Models can reach several gigabytes per training session. Choose a location on a drive with adequate free space — at minimum 10 GB, ideally 50 GB or more if you plan multiple training sessions.

1. Click Create Project.

The application creates the project folder structure, initializes the configuration file, and navigates you to the Split Dataset screen to begin preparing your data.

Opening an Existing Project

1. From the Home screen, click Browse Projects.

2. Your recent projects are listed with their name, path, and last opened date.

3. Click a project to open it. The application validates the project folder and loads the configuration.

4. If a project folder has been moved or deleted, you will see an option to Relocate it by pointing to the new location.

Editing a Project

From the Browse Projects screen, you can edit an existing project's name and coffee metadata:

1. Click the Edit button next to the project you want to modify.

2. Update the project name or any metadata fields.

3. Click Save.

Renaming a project also renames the folder on disk. All internal paths (checkpoints, datasets, exports) are updated automatically.

Deleting a Project

From the Browse Projects screen, you can remove a project:

1. Click the Delete button next to the project.

2. You will be asked whether to also delete the project files from disk (moved to your system's Recycle Bin) or only remove it from the project list.

3. Confirm your choice.