크게 세 가지가 가능한데
VOC dataset은 detection을 위한 것이고
structured folder는 classification
tfds는 classification과 detection 모두를 위한 것인데...
3.3.1 VOC dataset VOC datasets are introduced in Pascal Visual Object Classes Challenge. This format is also used by the ImageNet datasets. Images and labels in this dataset can be used to train detection models. |
3.3.2 Structured folders dataset Using the Import dataset screen, you can also import datasets that have a specific folder structure. These datasets do not have to be compressed into TAR files as with VOC datasets. The only requirement is that the images are saved in specific folders. The datasets created this way can be used to train models for classification tasks. |
3.3.3 TFDS datasets TFDS datasets allow you to import some of the datasets provided by tensorflow in their catalog here. The tool allows you to download datasets for image classification and object detection. |
그러면 segmentation은 누구냐..
문서 상에서는 cli를 통해서 직접해야 하는건가..
7.1 Image segmentation The following models target the image segmentation use case: • deeplab_v3 To run the “deeplab” model, the workspace\models\deeplab_v3\runner_demo.py script runs the inference using the converted model on the target device and decodes the output. You may also inspect the Python code. The script sends a simple HTTP request to the device and the rest of the script decodes the output. There are also a few images available for testing in the imgs folder. |
'embeded > i.mx 8m plus' 카테고리의 다른 글
ubuntu 22.04 + cuda + cudnn 설치 (0) | 2025.09.04 |
---|---|
import tensorflow illegal instruction (0) | 2025.09.04 |
eiq on windows with nvidia (0) | 2025.09.03 |
vainfo (0) | 2025.09.03 |
ubuntu nvidia 드라이버 설치 (0) | 2025.09.03 |