|
Greetings everyone! 👋🏻
In the August newsletter, our primary focus will be providing comprehensive insights into Nota AI's latest NetsPresso® lineup, LaunchX, and sharing updates about PyNetsPresso. Additionally, we're excited to announce that Nota AI's Hugging Face demo space has been selected as this week's 'Spaces of the week.'
Furthermore, you'll also discover details about Nota AI's recent patent registrations in both the United States and Japan, along with the signing of a Memorandum of Understanding (MOU) aimed at advancing mobility AI solutions. Be sure to stay tuned for these diverse achievements and updates! 👏🏻
Let's dive in together. 🌊
|
|
LaunchX is a product within the line of NetsPresso®, a hardware-aware AI model optimization platform, and it features the Model Launcher module as a standalone service. Let us tell you more about this product. 🙂
|
|
LaunchX is an independent service that enhances the functionality of NetsPresso®'s Model Launcher module. It streamlines the complex conversion process into two simple steps, enabling users to acquire benchmarks for various devices and identify the optimal device for their model. Additionally, it provides free models, ensuring accessibility even for those without an existing model.
|
|
|
The "Convert" feature of LaunchX streamlines the conversion process into two steps. All that is required is for users to upload their model and press the button. This automatically transforms their model into settings optimized for deployment on the target device, minimizing errors.
The following options are available for conversion!👍🏻
- ONNX → TensorRT
- ONNX → TensorFlow Lite
- ONNX → OpenVINO
- TensorFlow-Keras → TensorFlow Lite
|
|
|
The "Benchmark" feature of LaunchX allows you to assess your model's performance without the need for actual physical devices. During benchmarking, you can determine which hardware optimizes your desired performance without purchasing the physical hardware. Furthermore, it allows you to validate your target hardware's compatibility and AI model's compatibility!
The list of hardware available for benchmarking is as follows! 👐🏻
1. Raspberry Pi:
- Raspberry Pi Zero
- Raspberry Pi 3B+
- Raspberry Pi 2W
- Raspberry Pi 4B
2. NVIDIA Jetson Series:
- NVIDIA Jetson Nano
- NVIDIA Jetson TX2
- NVIDIA Jetson AGX Orin
- NVIDIA Jetson Xavier NX
- NVIDIA Jetson AGX Xavier
- NVIDIA Jetson AWS-T4
3. Intel Xeon:
Ready to optimize your model for your target device? Give LaunchX a try now! 🙌🏻
|
|
PyNetsPresso, an AI model development tool that employs Python code for optimizing models for target devices, has been updated. 👏🏻
This latest version offers new features such as the capability to transform AI models into different file formats using Python code. Also, a module named Launcher is now available, providing benchmarking capability. With this, users can measure the performance of their models on the target hardware. PyNetsPresso allows users to optimize and deploy their AI models onto target devices during development. Nota AI provides repositories on GitHub containing all the needed code for this tool. Users can utilize these resources to develop and improve their custom AI models into optimized ones. 😉
Streamline your AI model optimization workflow with PyNetsPresso!
|
|
🤗
Hugging Face picks Nota AI
for 'Spaces of the week'
|
|
|
Nota AI's Hugging Face demo space has been chosen as the 'Spaces of the week'! 👀 This space features two demonstrations that showcase Nota AI's advanced AI optimization technology.
The first demo features a lightweight Wav2Lip model. Users can create videos where a person's lip movements match the provided audio, using three videos and audio clips. By using selective quantization, this demo boosts edge GPUs' speed by up to 19 times without sacrificing quality.
The second demo showcases a streamlined Stable Diffusion model. Users can generate images for 20 scenarios, like "tropical birds on branches" or "floating colorful umbrellas." This demo reduces parameters and latency by 36% compared to prior models. It achieves high-quality generation by transferring knowledge from the original model, even with limited training resources.
These demos have been accepted at the 2023 ICCV (International Conference on Computer Vision), showing Nota AI's technical skills. To find out more, visit Nota AI's Hugging Face Demo Space now! 😎
|
|
✨
Nota AI secures US and Japan
patents for innovations
|
|
|
Nota AI's Research and Development (R&D) as well as Intellectual Property (IP) teams have achieved a significant milestone with the successful registration of patents in both the United States and Japan. 🙌🏻
We recently received a patent in the United States and Japan that covers 'Method for compressing neural network model and electronic apparatus for performing the same' and 'Method for lightweighting neural network model and electronic apparatus for performing the same' each.
These patents are utilized within NetsPresso®️, a hardware-aware AI model optimization platform, Model Compressor. This module offers users a user-friendly interface to analyze and compress AI models according to their desired compression ratios. By doing so, it enables users to achieve maximum performance and efficiency.
With 26 patents already granted and 90 patent applications domestically and globally. Stay tuned for how Nota AI's global technological capabilities will continue strengthening through ongoing domestic and international efforts toward further patent registrations! 🤗
|
|
🤝🏻
Nota AI, Telechips ink MOU
to advance mobility AI solutions
|
|
|
Nota AI and Telechips, a renowned international fabless semiconductor company, have signed a memorandum of understanding (MOU) to collaborate on creating AI solutions for the automotive industry! 🔥
This agreement involves both entities committing to collaboration and mutual support in developing high-performance AI models based on ADAS semiconductor chips and exploring new business opportunities.
Within this collaboration, Nota AI's AI solutions will be integrated into Telechips' semiconductor products, providing advanced AI solutions to NPU, semiconductor, and AI solution companies within the automotive sector. Subsequently, the companies plan to conduct customized prototype development and performance testing to meet customer requirements.
Through this partnership with Telechips, Nota AI aims to strengthen its position in the high-demand ADAS semiconductor-based AI solutions market. We are also dedicated to enhancing their product competitiveness, positioning themselves as a leading player in the mobility sector. 😉
|
|
📃
AI model optimization case
powered by NetsPresso®
|
|
|
|
Empowering Pedestrian Safety: The Obstacle Detection App and AI Model Optimization
|
|
Nota AI has developed a revolutionary application using NetsPresso®, a hardware-aware AI model optimization platform that detects obstacles. This app is specifically designed to recognize potential roadblocks and provide safety assistance to individuals who might struggle to perceive such obstacles. Users can receive timely warnings through voice or vibration alerts. 🚶🏻♂️
The obstacle detection application employs advanced AI object recognition technology to analyze real-time camera input from users, enabling it to identify potential barriers along the road. To learn more, click the button below and read the article posted in the Community tab on Nota AI's official website.
|
|
🗣️
Upcoming events in September
|
|
|
|
2023 World Smart City Expo
|
|
Nota AI will be at the World Smart City Expo from September 6th to 8th at KINTEX Exhibition Center 1 in Ilsan, Republic of Korea. We will be showcasing our its Nota ITS solution, which utilizes AI optimization technology, with its primary focus being the presentation of AI Safe Crossing solution and associated case studies. 🚦
|
|
Nota AI will participate in TechCrunch Disrupt 2023 in San Francisco, USA, from September 19th to 21st (local time)!
Nota AI will be demonstrating its NetsPresso® platform, including pothole detection and laundry symbol recognition Android apps developed by the NetsPresso Application team as well as use case videos. 📹🙌🏻
|
|
|
|
