How to Generate Projects by Importing Datasets Using the eIQ Toolkit (eIQ Portal)

[masayuki_igarashi]

After installing the eIQ Toolkit, I attempted to generate a project by importing a dataset, but no images appeared on the Data Set Curator screen.
Please review the installation steps and eIQ Toolkit operation procedures below. Could you identify any missing or incorrect steps?
Additionally, if any tools are required to run eIQ Toolkit on Ubuntu, please advise.

■Environment
OS: Ubuntu 20.04.6 (on VMware running Windows 11)
eIQ Toolkit version: 1.16.0
Dataset:
　　- Train Group: VOCtrainval_06-Nov-2007.tar,
　　- Test Group: VOCtest_06-Nov-2007.tar

■Issue
From the eIQ Portal - HOME screen, I select Create Project -> import datasets and specify the files listed in the “Environment” section above for the Train Group and Test Group. After clicking the IMPORT button, the Data Set Curator screen appears, but it displays “no images found.” and no images are shown.

■Procedure
1. Install the eIQ Toolkit
2. From the application display screen (where you can search for applications), search for “eIQ”. The eIQ Toolkit icon will appear; click it to launch the application.
3. On the application's HOME screen, click import dataset in the CREATE PROJECT dropdown menu.
4. The Import Datasets screen appears on the right side. Click VOC-Dataset on that screen. This displays the Train group: and Test group: dataset selection screen on the left side.
5. Specify “VOCtrainval_06-Nov-2007.tar” for Train group: and “VOCtest_06-Nov-2007.tar” for Test group:.
6. Click the IMPORT button. You can then specify the storage location and name for the generated project. Save it in any directory with the name “eiq-project.eiqp”.
7. After “Loading Test dataset” and “Loading train dataset” appear to complete, the Data Set Curator screen displays on the right side. It shows “No images found.” Additionally, the left side shows “Unlabeled images” as 0 (zero), and “All Images” displays three zeros: 0, 0, 0.

■Other Verifications Checked
1. checked the presence of image files within the datasets and their display capability. Both datasets (VOCtrainval_06-Nov-2007.tar and VOCtest_06-Nov-2007.tar) could be extracted. After extraction, image files (e.g., VOCtrainval_06-Nov-2007/VOCdevkit/VOC2007/JPEGImages/000005.jpg) were confirmed to exist, and double-clicking these files opened the images in an image viewer application.
2. On the Windows version of eIQ Toolkit, I confirmed images display in the Data Set Curator screen.
3. I launched eIQ Toolkit (eIQ Portal) using the following command instead of the icon, but the result was the same:
$ source /opt/nxp/eIQ_Toolkit_v1.16.0/bin/eiqenv.sh
$ /opt/nxp/eIQ_Toolkit_v1.16.0/eiq-portal

■Attached Documents
1. Log when launching eIQ Toolkit via command: exe_log_20251017_083542.log
2. Log capture during eIQ Toolkit installation (Japanese environment): install.png
3. Data Set Curator screen: data_set_curator.png

[Gavin_Jia · Accepted solution]

Hi @masayuki_igarashi ,

Thanks for your interest in NXP MIMXRT series!

I tested it locally and it displayed fine after importing the dataset. Although I'm using Windows, the process should be the same. The problem may be that the folder structure and the annotations are not aligned.

Here are my test environment and results:

- Windows11
- eIQ 1.16

Result:

I couldn't get the original dataset, so I downloaded it from this page. You can also try it. Simply unzip the downloaded *tar.zip file and find the *.tar file. Then, select this *.tar file when importing into eIQ.

Best regards,
Gavin

[masayuki_igarashi]

Dear, @Gavin_Jia ,

We have received the new data and can now display the images.
Following that, we have been able to proceed with training and model creation.
Thank you very much

Best regards, Masayuki Igarashi.