Open-source alternatives to Jetson Inference

ExecuTorch is a lightweight C++ runtime for deploying PyTorch models on mobile, embedded, and edge hardware. It provides an ahead-of-time compilation pipeline that exports, quantizes, and lowers model graphs into compact serialized programs, then executes them through a minimal runtime with hardware acceleration and on-device large language model inference capabilities. The project distinguishes itself through a hardware accelerator delegate system that partitions model subgraphs and offloads computation to specialized backends including NPUs, GPUs, and DSPs from Apple, Arm, Intel, MediaTek,

GoCV is a computer vision library and Go language binding for OpenCV. It serves as an image processing toolkit and deep learning inference engine, providing programmatic access to a wide range of algorithms for image manipulation, object detection, and video analysis. The project differentiates itself through high-performance native bindings and hardware acceleration. It utilizes a foreign function interface to map Go calls to C++ functions and includes a hardware-agnostic backend dispatch to route neural network tasks to computation engines such as CUDA and OpenVINO. The library covers a br

This project is a comprehensive collection of educational examples and reference implementations for building vision and language models using PyTorch. It serves as a deep learning tutorial covering the end-to-end process of developing neural networks, from initial architecture definition to final production deployment. The repository provides detailed guides on implementing a wide range of domain-specific models, including convolutional neural networks for object detection and segmentation, as well as transformer and recurrent architectures for natural language processing. It emphasizes gene

MNN is a high-performance inference engine and framework designed for on-device machine learning. It provides a comprehensive environment for executing, optimizing, and deploying neural network models directly on mobile and resource-constrained edge devices. The framework distinguishes itself through a robust model optimization toolkit that supports quantization, compression, and structural graph manipulation to minimize memory footprint and maximize execution speed. It features a modular architecture that abstracts hardware-specific backends, allowing models to run efficiently across diverse

ncnn is a high-performance neural network inference framework designed for executing deep learning models locally on mobile and desktop hardware. It functions as a specialized engine that enables the deployment of artificial intelligence tasks directly on resource-constrained devices, eliminating the need for external network connectivity or cloud-based processing services. The framework provides a comprehensive toolset for model optimization, allowing users to convert and quantize machine learning models into specialized binary structures. By utilizing static model graph compilation and zero

This project provides a comprehensive technical guide and framework for engineering large-scale machine learning systems. It covers the full lifecycle of model development, focusing on the infrastructure and computational principles required to build, train, and serve generative AI models across distributed GPU clusters. The repository distinguishes itself by offering deep-dive tutorials and implementation strategies for complex system challenges. It emphasizes high-performance architectural primitives, such as collective communication orchestration, distributed tensor sharding, and static gr

Sglang is a high-performance inference engine and serving system designed for large language and multimodal models. It provides a programmable interface for orchestrating complex generation workflows, enabling developers to coordinate multi-turn dialogues, tool invocations, and reasoning chains through a domain-specific language. The platform is built to support production-scale deployments, offering an OpenAI-compatible API that allows for integration with existing application ecosystems. The system distinguishes itself through a disaggregated architecture that separates compute-intensive pr

lite.ai.toolkit is a C++ computer vision toolkit designed for edge AI deployment. It enables the execution of pre-trained models for object detection, image classification, and segmentation on resource-constrained devices. The project features a multi-backend inference engine that supports the ONNX model runtime, allowing AI models to run across different hardware targets. It includes a GPU-accelerated pipeline specifically for NVIDIA hardware to reduce latency and increase processing speed. The toolkit covers a broad range of facial analysis capabilities, including emotion detection, gender

AISystem is a comprehensive AI full-stack infrastructure project covering the entire pipeline from AI chip architecture to high-level training frameworks. It encompasses the development of AI compiler frameworks, inference engines, and distributed training orchestrators designed to coordinate workloads across a heterogeneous compute stack of CPUs, GPUs, and NPUs. The project focuses on the deep integration of software and hardware, employing software-hardware co-design to align tensor layouts with physical memory structures. It provides specialized capabilities for accelerating Transformer mo

Paddle-Lite is a deep learning inference engine and edge computing runtime designed to execute trained models on mobile and edge devices. It provides a hardware-accelerated inference framework and a decoupled runtime with a minimal binary footprint to operate in resource-constrained environments without third-party dependencies. The project includes a model quantization tool for reducing precision and size via static and dynamic quantization, as well as a computation graph optimizer. These tools reduce latency and memory usage by fusing operators and pruning the model intermediate representat

waifu2x-caffe is a deep learning image upscaler and denoiser that uses the Caffe framework to increase image resolution and remove noise from illustrations and photographs. It functions as a neural network image processor that reduces compression artifacts and pixelation while maintaining visual clarity. The project provides specialized neural network weights optimized separately for 2D illustrations and real-world photographs. It includes distinct processing for alpha channels to preserve transparency and employs test-time augmentation to improve output precision. The tool supports both a c

TengineKit is a mobile computer vision software development kit designed for real-time inference on local hardware. It functions as a neural network engine that executes deep learning models directly on mobile devices, enabling applications to perform complex visual analysis without relying on cloud connectivity. The framework provides specialized tools for detecting and tracking human features, including faces, hands, bodies, and irises, alongside general object detection capabilities. By utilizing a native core runtime and hardware-accelerated execution, the library processes visual data lo

Neutts is a neural text-to-speech engine designed for real-time streaming output on edge devices such as phones and laptops. It supports voice cloning from short audio references, enabling zero-shot reproduction of a target speaker's voice, and can be fine-tuned or retrained from scratch for custom voices and styles. The system distinguishes itself through a decoder-only architecture that halves memory and accelerates generation on constrained hardware, combined with quantized model inference for reduced memory footprint. Its streaming decoder loop interleaves synthesis with playback, deliver

FlashMLA is an LLM attention kernel library and inference acceleration library providing a collection of high-performance CUDA kernels. It implements multi-head latent attention mechanisms designed to reduce memory overhead and increase throughput during the forward and backward passes of large language model inference. The library utilizes quantized cache attention kernels to improve computation efficiency across both sparse and dense token processing. It specifically optimizes the prefill and decoding phases of model inference through these latent attention implementations. The project cov

Mace is a mobile deep learning inference framework and hardware acceleration engine. It functions as a runtime for executing neural network models on mobile devices, distributing computations across CPUs, GPUs, and NPUs. The project includes a cross-platform model converter for transforming pre-trained neural networks from various industry formats into mobile-optimized representations. It also provides a neural network obfuscator that converts model weights into source code to protect intellectual property from reverse engineering. The framework manages on-device resources by optimizing memo

This project is a PyTorch object detection framework that implements the Faster R-CNN architecture. It serves as a vision model for predicting precise bounding boxes around multiple objects within images and live video feeds. The system is optimized for multi-GPU training to reduce the time required for model convergence. It utilizes a GPU-accelerated design to handle the training and inference of complex detection networks. The framework covers the full object detection lifecycle, including custom network training and inference for static images and real-time video streams. It includes capa

This project provides a foundational framework and reference implementation for executing causal language modeling and multimodal reasoning on local systems. It includes a set of core components for managing model assets, a fine-tuning framework, and structural definitions required to instantiate transformer-based architectures. The system is distinguished by its ability to process combined text and image inputs through multimodal transformer models for visual reasoning and document analysis. It also supports the deployment of quantized models, reducing memory footprints through low-precision

TensorRT is a deep learning inference engine and software development kit designed to optimize and deploy neural networks for high-performance execution on NVIDIA GPUs. It functions as a GPU acceleration framework that reduces latency and increases throughput for trained models during production deployment. The toolkit imports models from the Open Neural Network Exchange format and transforms them into optimized engines. It utilizes graph-based model optimization, layer-fusion kernel generation, and precision-based quantization to convert floating point weights into lower precision formats.

Darkflow is an object detection framework and computer vision pipeline that provides a programmatic interface for performing real-time image analysis and object identification. It functions as a tool for loading weights, fine-tuning models, and executing inference on both static images and video feeds. The project serves as a converter that translates Darknet configurations and weights into TensorFlow graphs to enable retraining and deployment. It includes a model exporter that saves trained graphs into portable protobuf files for use on mobile and native devices. The system covers capabilit

DeepFaceLive is a desktop application designed for real-time facial replacement and animation within live video streams. By utilizing deep learning models, the software performs high-speed identity mapping and facial feature analysis to transform video content as it is captured. The engine relies on GPU-accelerated inference to execute these complex image manipulation tasks at interactive frame rates. The application distinguishes itself through a modular video processing pipeline that chains specialized tasks to maintain high throughput and low latency. It features a virtual camera streaming

oneDNN is a library for deep learning acceleration that provides optimized building blocks for neural network training and inference. It manages tensor computation across CPU and GPU hardware, enabling the execution of high-performance primitives for model training and neural network inference optimization. The project distinguishes itself through hardware-specific kernel optimization and the use of just-in-time compilation to target specific processor instruction sets. It supports quantized neural network execution using both static and dynamic quantization to reduce memory usage and increas

OpenNMT-py is a PyTorch neural machine translation framework used for training and deploying neural machine translation and large language models. It functions as a distributed model training system, an inference engine, and a toolkit for fine-tuning large language models. The framework distinguishes itself with a dedicated toolkit for adapting large language models through low-rank adaptation, quantization, and instruction tuning. It also includes a neural machine translation server that allows trained models to be hosted and exposed via REST API endpoints. The project covers a broad range

NanoKVM is a KVM-over-IP device that provides remote keyboard, video, and mouse control over IP networks for headless server management. It functions as remote server management hardware enabling out-of-band control of a computer's power, BIOS, and operating system over a network, while also serving as a RISC-V single-board computer for embedded and edge applications. The device additionally operates as an AI edge inference device running neural network models locally for real-time image recognition and object detection, and integrates Tailscale as a VPN appliance for secure peer-to-peer conne

ImageAI is a Python computer vision library providing a suite of tools for image classification, object detection, and video analytics. It functions as an integrated framework for locating and labeling objects in static images and video streams, utilizing deep learning models for identification and categorization. The project includes a model training toolkit that allows for the creation of custom classifiers and detectors through scratch training or transfer learning. It features a GPU-accelerated inference engine to increase processing speed for vision tasks and includes specialized utiliti

LAVFilters is an open-source media filter pack consisting of splitters and decoders designed for the DirectShow framework. It functions as a set of software components that convert compressed media data into raw formats for hardware or software rendering. The collection includes a media splitter to separate combined audio and video container formats into individual streams, as well as a specialized Blu-ray disc player filter used to identify and extract movie titles and playlists from disc structures. The project provides capabilities for audio and video stream decoding, media stream demuxin

Human is a TensorFlow.js computer vision library used for face, body, and hand tracking within the browser or Node.js. It provides a framework for human pose and gesture tracking, facial recognition, and biometric liveness detection to verify a live human presence. The project distinguishes itself through a full suite of identity and motion tools, including a facial recognition framework that generates embeddings for similarity matching and a background segmenter for separating humans from their environment. It incorporates a liveness detector to prevent spoofing during facial analysis. The

LMCache is a distributed key-value cache manager and tiering system designed to accelerate large language model inference. It functions as a tiered storage layer that offloads tensors from GPU memory to CPU RAM, local disks, or remote object stores, enabling the reuse of cached prefixes across different inference sessions and serving engines. The system differentiates itself through a disaggregated prefill-decode model, which separates prompt processing from token generation by transferring caches between distributed compute nodes. It utilizes peer-to-peer orchestration to share and retrieve