Understand the deprecation of NVIDIA TAO models and choose supported Edge Impulse alternatives.
Deprecated featureThis feature has been deprecated. Please see below for additional details.
Edge Impulse support of NVIDIA TAO object detection and image classification models has been deprecated. As of February 28th, 2025, these model options no longer appear within the platform.
Your existing projects that use NVIDIA TAO models will continue to work in Studio, including deployment. However, you will not be able retrain these models.
There are several object detection models built into the platform, including those based on the Edge Impulse YOLO-Pro and FOMO architectures, in addition to the MobileNetV2 SSD FPN architecture. These models can be accessed by adding an object detection learning block to your impulse, then selecting your desired model on the settings page for the block.You are also able to train other object detection architectures using custom learning blocks. For example, see this YOLOv5 learning block.All of these options can be used as alternatives to the NVIDIA TAO object detection models that were previously available.
In addition to existing models available for image classification found within the platform, the EfficientNet architecture has now been added as well. These models can be accessed by adding a transfer learning (images) learning block to your impulse, then selecting your desired model on the settings page for the block.Similar to the object detection section above, other image classification architectures can be trained using custom learning blocks.All of these options can be used as alternatives to the NVIDIA TAO image classification models that were previously available.