The iCore developer tools include everything you need to create edge AI systems.
The Intellino Simulator provides two operation modes:
① Existing Datasets Mode
Allows users to easily perform training and inference using pre-prepared datasets.
② Custom Datasets Mode
Allows users to manually configure datasets and training parameters,
and perform training and inference according to the Intellino board’s memory size (2K · 8K · 16K).
Up to 5 experiments with different parameter combinations can be compared for accuracy evaluation.
How to Use Existing Datasets Mode
Select the dataset you want to use.
Choose the image or data file you want to run inference on.
Press the Start button to execute inference, and the result will be displayed on the screen.
You may choose other images and perform inference repeatedly.
How to Use Custom Dataset Mode
1. Setting Memory Size and Parameters
Select Custom Dataset Mode.
Choose a board memory size from 2K / 8K / 16K.
Enter the following three training parameters:
Input Vector Length (V)
Number of Classes to Train (C)
Number of Training Datasets per Class (T)
The following condition must be satisfied to proceed:
If the condition is not satisfied, the Next button remains disabled.
After entering the parameters, press the Apply button to confirm the settings.
The Next button becomes active only after pressing Apply button.
2. Selecting the Training Dataset Folder
Press the Next button to move to the folder selection screen.
Select the datasets folder to be used for training
(any local folder can be used).
The folder must satisfy the following conditions:
Number of class folders = entered value C
Number of images per class = entered value T
If the structure does not match the parameters, training cannot proceed correctly.
3. Running Inference
On the inference screen, select the image file you want to test.
Press the Start button to execute inference.
The result will be displayed on the screen, and you may select other images
to perform inference repeatedly.
4. Accuracy Comparison Feature
After completing inference, press the Next button to move to the accuracy comparison screen.
The accuracy graph of all experiments performed so far will be displayed.
Each experiment record includes the following parameters:
V: Input Vector Length
C: Number of Classes to Train
T: Number of Training Datasets per Class
Reconfigure Button
Allows you to re-set the memory size and parameters (V, C, T)
and add a new experiment.
Up to 5 experiments with different parameter combinations can be compared.
Once all 5 slots are filled, the Reconfigure button becomes disabled.
5. Finishing and Resetting Experiments
Pressing the Finish button clears all stored accuracy records.
After resetting, you can start new experiments again from the initial parameter-setting step.