60 docs tagged with "MPU"

Build a Q&A assistant that picks answers using on-device embeddings.

Let an on-device AI model control an LED and system functions via tool calling.

Measure the inference speed of pre-trained YOLO models on Astra.

Compile your own model with the SyNAP toolkit and run it on the SL1600 NPU.

Convert your own model with the Torq compiler and run it on the SL2600 NPU.

Build llama-cpp-python on the board for local LLM inference from Python.

A preprocessing model to convert NV12 formatted images from 1920x1080 resolution to RGB format at the same resolution.

A preprocessing model to convert NV12 formatted images from 1920x1080 resolution to RGB format at 224x224 resolution.

A preprocessing model to convert NV12 formatted images from 1920x1080 resolution to RGB format at 640x360 resolution.

Cross-compile llama.cpp on a host machine for Astra Machina.

Learn how text embeddings work and generate them on-device with Synaptics Astra.

A three-stage on-device face recognition pipeline: face detection, 68-point facial keypoints, and 256-D face embeddings.

A compact 270M parameter multimodal language model from Google optimized for text classification and data extraction tasks with rapid fine-tuning capability.

A compact 270M parameter language model from Google, quantized to INT4, for text classification and data extraction tasks with rapid fine-tuning capability.

Set up your board and run your first edge AI examples.

Set up your board and run your first edge AI examples.

Recognize and label images with a classifier on the SL1600 NPU.

Recognize and label images with a classifier on the SL2600 NPU.

A quantized Inception V4 model optimized for image classification on ImageNet at 299x299 resolution.

Run small large language models locally on Astra.

A 450M parameter vision-language model from Liquid AI that answers natural-language questions about an input image.

A 230M parameter hybrid short-convolution and attention language model from Liquid AI that can act as a skill-selection layer, decomposing instructions into tool calls.

Run Llama 3.2 1B on Astra as a single llamafile executable.

Run TinyLlama natively on Astra Machina with llama.cpp.

A MobileNet V1 model with a 0.25 width multiplier optimized for image classification on ImageNet at 224x224 resolution.

A quantized MobileNet V1 model with a 0.25 width multiplier optimized for image classification on ImageNet at 224x224 resolution.

A MobileNet V2 model with a 1.0 width multiplier optimized for image classification on ImageNet at 224x224 resolution.

A quantized MobileNet V2 model with a 1.0 width multiplier optimized for image classification on ImageNet at 224x224 resolution.

A quantized MobileNet V2 model with a 1.0 width multiplier optimized for image classification on ImageNet at 224x224 resolution.

A lightweight MobileNet model optimized for people detection on high-resolution images.

A MobileNet model optimized for object detection on the COCO dataset with full resolution at 224x224 and 80 classes.

Modify the SyNAP framework to build your own C++ inference apps.

A streaming automatic speech recognition model with a sliding-window encoder for low time-to-first-token, optimized for 16kHz mono audio.

A high-efficiency automatic speech recognition model using variable-length architecture for real-time speech recognition, optimized for 16kHz mono audio.

Build C++ apps that run models on the NPU with the SyNAP framework.

Find and locate objects in images with YOLO on the SL1600 NPU.

Find and locate objects in images with YOLO on the SL2600 NPU.

A PoseNet model using MobileNet architecture with 75% width multiplier for efficient body pose estimation.

A quantized PoseNet model using MobileNet architecture with 75% width multiplier for efficient and optimized body pose estimation.

Fine-tune YOLO11 for PPE detection with Roboflow and deploy it to Astra.

Transcribe speech on-device with Whisper and Moonshine.

A fast super-resolution model converting YUV images from 1280x720 to 3840x2160 resolution.

A fast super-resolution model converting YUV images from 1920x1080 to 3840x2160 resolution.

A QDEO-based super-resolution model converting YUV images from 1280x720 to 3840x2160 resolution.

A QDEO-based super-resolution model converting YUV images from 1920x1080 to 3840x2160 resolution.

Turn written text into natural-sounding speech on-device with Piper.

An overview of the Torq Edge AI platform for running neural networks on the SL2610 NPU.

Export an Ultralytics YOLOv8 pose model and compile it for the Astra NPU.

Run vision language models that understand images and text with llama.cpp.

Transcribe a WAV file on-device with whisper.cpp.

Transcribe live microphone audio in real time with whisper.cpp.

YOLO26 NMS-free object detection models (nano and small) optimized for 320x320 resolution, with the decode head compiled onto the NPU.

A YOLOv5m model for object detection optimized for 640x480 resolution, offering a balance between speed and accuracy.

A YOLOv5s model for object detection optimized for 640x480 resolution, suitable for real-time applications.

A YOLOv5s model specialized for face detection, optimized for 640x480 resolution using ONNX with mixed quantization.

A YOLOv8n model for object detection optimized for 320x320 resolution, offering a balance between speed and accuracy.

A YOLOv8n pose estimation model optimized for 320x320 resolution, detecting people and their body keypoints.

A YOLOv8s model specialized for body pose estimation, optimized for real-time applications.

A YOLOv8s pose estimation model optimized for 320x320 resolution, detecting people and their body keypoints.

A YOLOv8s model for object detection optimized for 320x320 resolution, offering a balance between speed and accuracy.