How It Works
ODHUB will automatically validate the accuracy and sufficiency of the data, which includes image files and XML label files. ODHUB also provides a built-in labeling tool, which allows users to perform the labeling task if necessary
Product Features
No coding
ODHUB is specifically designed for engineers who do not have a profound knowledge of programming skills to design a proper object detection model in only a few simple clicks. It provides a graphical user interface (GUI) that makes it easy to select the appropriate components, configure the model, and train it.​​​
.png)
Built-in labeling tool
ODHUB's built-in labeling tool enables users to add, modify, or delete labels in their training datasets. The labeled files are saved in PASCAL VOC format, a widely used format for object detection datasets. The labeling tool supports bounding box and polygon labeling types.
-
Bounding box: This is the most common type of labeling. It involves drawing a rectangle around the object in the image. Bounding boxes provide a good approximation of object position and size but may not accurately capture intricate object contours or irregular shapes.
-
Polygon: This is used for objects that cannot be easily represented by a rectangle, such as people or animals. It involves drawing a polygon around the object. Polygon annotations offer more precise boundaries, ensuring higher accuracy in capturing complex object shapes and contours.
Advanced frameworks
ODHUB supports both TensorFlow and YOLO frameworks for object detection. TensorFlow frameworks include SSD, Faster RCNN, and EfficientDet, while YOLO frameworks include YOLO2, YOLO3, and YOLO4. This allows ODHUB to achieve high inference speed and accuracy.
ODHUB uses NVIDIA GPU to speed up the training process. It can also process data locally to ensure data privacy. The training parameters are easily configured, making it a convenient tool for object detection.
Model evaluation & testing
ODHUB's built-in evaluation and testing interface allows users to test the performance of the models before final deployment. Users can select any image dataset or built-in video capture devices in their system to test how the trained model performs object detection tasks in real time.
Optimized for LabVIEW
ODHUB allows you to export your models in .zip format. The zip file contains all the necessary information, such as the model's architecture, weights, and biases, to deploy in LabVIEW using ANSDL
Optimized for hardware
ODHUB allows developers to export object detection models in optimized formats that can be directly used on different hardware using ANSVIS. This offers great opportunities for developers to bring AI to the Internet of Things (IoT) field.
