Nexa SDK is a comprehensive toolkit for supporting GGML and ONNX models. It supports text generation, image generation, vision-language models (VLM), Audio Language Model, auto-speech-recognition (ASR), and text-to-speech (TTS) capabilities.

Multi-Task Temporal Shift Attention Networks for On-Device Contactless Vitals Measurement (NeurIPS 2020)

This repository contains notebooks that show the usage of TensorFlow Lite for quantizing deep neural networks.

A custom RAG pipeline for multi-document QA from PDF/DOCX documents, in Android

This is a web demo for camera-based PPG sensing (rPPG).

An Android app running inference on Depth-Anything and Depth-Anything-V2

Embeddings from sentence-transformers in Android! Supports all-MiniLM-L6-V2, bge-small-en, snowflake-arctic, model2vec models and more

Object detection inference with Roboflow Train models on NVIDIA Jetson devices.

Approach to implementing distributed training of an ML model: server/device training for iOS.

[NeurIPS'24] DEX: Data Channel Extension for Efficient CNN Inference on Tiny AI Accelerators

End-to-end on-device federated learning, "An On-Device Federated Learning System for SMS Spam Classification", IEEE MIT URTC 2022

SponsorMe is a project to help provide access to digital tools for learning, powered by on-device machine learning, innovation and willingness, to those people that have limited access to technology due to demographics, disabilities, economy or other multiple reasons.

As described in "Towards Full On-Tiny-Device Learning: Guided Search for a Randomly Initialized Neural Network"

Python ML library for person fall detection. Intended for IoT deployments with on-device inference and on-device transfer learning.

This repository is meant to be a playground for mobile deep learning.

Control your computer using hand gestures with AI, using Google's MediaPipe and OAK-D Lite camera.

*Resource Efficient Federated Deep Learning for IoT Security Monitoring

An Android app where users draw a number and machine learning does the rest

A minimalistic Android app showcasing semantic search using ObjectBox and Lucene KNN, leveraging the MiniLM-L6-V2 embedding model and bert_vocab.txt for efficient retrieval.