Fast, Lightweight and On-Device: How Samsung Research Built AI Features that Translate in Real Time
Our real-time Interpreter and Live Translate feature based on on-device AI expands its list of supported languages to 16. The on-device AI can provide these services without relying on external servers or the cloud, so users can be assured that their data remains private and secure.
Galaxy's AI translation feature available on various applications is an innovative technology developed by Samsung Research, who have accumulated translation data over a long time, integrated it with AI technology, and advanced it with the MX Business R&D office. Samsung Research’s Global AI Center helped to commercialize the on-device AI model by adding its proprietary technology to the self-developed AI translation model. Hear from Samsung Research Global AI Center's researchers leading the on-device AI field in the Samsung Electronics Newsroom.
|
|
|
|
Tutorial: Connect Galaxy Watch to Android Studio over Wi-Fi
Experience the convenience of launching, testing, and debugging applications on your watch wirelessly. You can run and test wearable applications wirelessly on Galaxy Watch4 or any later model through Android Studio. This tutorial walks you through how to wirelessly test and debug applications through an Android Debug Bridge (ADB) connection. Learn how to pair and connect a Galaxy Watch to Android Studio over Wi-Fi.
|
|
|
|
Tutorial: Optimizing Watch Face Battery Usage by Reducing On-Pixel Ratio
Watch faces can consume significant battery power when they are constantly in use, so it is important to optimize your watch face’s performance. This tutorial demonstrates how to optimize a watch face created in Watch Face Studio by reducing the On-Pixel Ratio (OPR). Check out the design tips for optimizing battery use.
|
|
|
|
MELS-TTS: Multi-Emotion, Multi-Lingual, Multi-Speaker Text-To-Speech System via Disentangled Style Tokens
As the text-to-speech (TTS) system technology based on neural networks progresses rapidly, the quest for creating human-like speech has taken remarkable strides forward. Recent advancements have opened avenues for TTS systems capable of not only mimicking human speech but also encapsulating the nuances of emotions and linguistic diversity.
With a growing demand for more sophisticated TTS capabilities, the pursuit of multi-emotion and multi-lingual TTS systems has become more important. However, this journey is riddled with complexities as it’s difficult to obtain speech samples from target speakers exhibiting multiple emotions or languages, and also to separate different speech attributes including content, speaker identity, emotional language, etc. MELS-TTS, proposed to address these issues, uses disentangled style tokens that separate many different speech attributes. Learn about the research on MELS-TTS and how it has proved to be superior to other reference-based TTS systems in multiple assessment scenarios.
Learn more |
|
|
|
FFT-Based Selection and Optimization of Statistics for Robust Recognition of Severely Corrupted Images
In recent years, home robot technologies have seen a huge improvement thanks to the progress of vision systems installed on them. Technologies such as object detection and recognition have helped robotic devices to avoid obstacles and recognize objects in their vicinity. However, these systems are prone to failure when they face challenging situations like light changes or adverse weather conditions.
To implement a robust vision model in a real-world application, it is imperative to address these challenges. To that end, this post introduces a novel, FFT-based RObust STatistics selection method (FROST). Learn more about FROST, including its key steps and characteristics, on the Samsung Research blog.
Learn more |
|
|
|
Signature-based Approach towards Global Channel Charting with Ultra Low Complexity
Channel charting (CC) is an unsupervised learning method that uses channel information to extract a low-dimensional embedding that preserves the geometrical structure of a physical space of user equipment. It can be used in a broad range of application areas such as handover, indoor localization, beam management, and so on. However, many past methods have been problematic as they mainly focused on charting that only preserved local geometry and used raw channel information without considering the global geometry, which meant that they were computationally intensive and time-consuming.
This article proposes a new, signature-based approach that generates a global chart with ultra-low complexity. Learn more about how this new approach helps overcome the constraints of existing methods and achieve better performance on the Samsung Research blog.
Learn more |
|
|
|
|
|
|
|
https://developer.samsung.com
Copyright© %%xtyear%% SAMSUNG All Rights Reserved.
This email was sent to %%emailaddr%% by Samsung Electronics Co.,Ltd.
You are receiving this email because you have subscribed to the Samsung Developer Newsletter through the website.
Samsung Electronics · 129 Samsung-ro · Yeongtong-gu · Suwon-si, Gyeonggi-do 16677 · South Korea
Privacy Policy
Unsubscribe
|
|
|
|
|
|
|
|
|
