News › ai-infrastructure
By Zayden R., August 22, 2026
The open-source Etnaviv driver can now run YOLOX object detection on Vivante NPUs, enhancing its capabilities beyond SSDLite MobileDet. This advancement is crucial for engineers using the NXP i.MX8M Plus.
The open-source Etnaviv driver has achieved a significant milestone by enabling support for YOLOX object detection on Vivante NPUs. This breakthrough, led by developer Tomeu Vizoso, expands the driver's capabilities beyond the previously supported SSDLite MobileDet model. For engineers utilizing the NXP i.MX8M Plus, this development means more robust and flexible AI processing capabilities.
Etnaviv, a reverse-engineered driver initially developed to provide accelerated graphics support for Vivante GPU IP, has broadened its scope to encompass Vivante NPUs as well. The recent integration of YOLOX, a more advanced object detection model, into the driver stack is a testament to the ongoing efforts to enhance open-source AI infrastructure.
The changes are part of a broader effort involving Mesa's Teflon framework, a project that has been instrumental in supporting AI and NPU functionalities. With the latest updates, the Etnaviv driver is now well-positioned as a flagship model for running complex AI tasks on Vivante hardware.
For those interested in exploring this further, Vizoso's blog provides additional insights into how YOLOX operates within the Etnaviv ecosystem. This development not only highlights the potential of open-source solutions in AI but also underscores the importance of community-driven innovation in expanding hardware capabilities.
Engineers working with Vivante NPUs should consider testing YOLOX on their systems to evaluate performance and integration. As the Etnaviv driver continues to evolve, it promises to offer even more sophisticated support for AI applications in the future.
The Linux Camp teaches these topics as hands-on labs on real virtual machines, verified as you type.