I’m reaching out to NVIDIA for support on behalf of our community. Recently, we’ve discussed adopting a more permissive license for YOLOv9. This decision was made with the hope that major companies like NVIDIA could contribute to an open project that benefits both the community and the broader public.
With this new implementation, we have the opportunity to leverage NVIDIA tools for training such as DALI and the TAO Toolkit, incorporating some of the best detection models available.
I have been developing several projects, including Triton Server, DeepStream, and TensorRT QAT , using YOLO models. Thanks to the more permissive license, we can now integrate the latest YOLOv9 models into the TAO Toolkit.
We are excited about the potential collaborations and advancements this can bring, and we look forward to NVIDIA’s support in enhancing our projects and community initiatives.
The latest models bring significant innovations, such as being NMS-free, which eliminates the need for Non-Maximum Suppression as post-processing. This, along with other advantages, greatly enhances detection and segmentation models.