Release 2024/11/12
Last updated
Last updated
We’re thrilled to introduce the latest updates in DataVerse, aimed at enhancing model conversion, data import options, and the annotation process. Here’s what’s new in November:
Model Convert (Beta):
The new model conversion feature allows you to convert models into ONNX and TensorRT formats, with options to quantize to FP16 or INT8. This improves detection efficiency, making models more suitable for real-time applications. Additionally, you can export converted models directly into Observ for seamless integration. We will continue expanding support for more environments.
YOLO Format Data Import:
DataVerse now supports data import in YOLO format, simplifying workflows for users who utilize YOLO-based datasets.
Training Progress Display:
A new progress display for training processes has been added, providing visibility into training status and helping you better manage your model development.
Enhanced Annotation Labeling Panel:
We’ve introduced several powerful tools in the annotation panel to streamline labeling workflows:
Comment System: Add and manage comments directly within annotations for better labelers / reviewers collaboration and feedback.
Sequence Batch Editor: Edit multiple sequences at once (paste, delete), making it faster to apply consistent changes.
Additional hotkeys: New keyboard shortcuts have been added for quicker, more efficient labeling.
Multi-Select and Copy-Paste: You can now select multiple objects, copy, and paste them, making it easier to manage repeated annotations.
Annotation Opacity: You can adjust object opacity to view your target annotations.
VQA (Visual Question Answering) Data Format:
We’re working on supporting the VQA data format, which will include visualization and annotation capabilities. This feature will allow for advanced question-and-answer-based annotations, enhancing the versatility of your datasets.
Be sure to follow us for more details and updates as we approach the release date.
For technical support, please contact our support team during business hours.
