|
Greetings everyone! 👋🏻
In this July newsletter, we're excited to announce that Nota AI has been accepted into the NVIDIA Cloud Validated Ecosystem, and we'd like to introduce you to NetsPresso® and the new lineup we've recently launched based on it.
Also, in this issue, we are thrilled to announce that Nota AI has earned a spot in the international top 10 in two divisions of the VOTS2023 Challenge!
Let's dive in together. 🌊
|
|
🌐
NVIDIA Cloud Validated Ecosystem
|
|
|
Nota AI is thrilled to announce that NetsPresso® has been registered in the NVIDIA Cloud Validated ecosystem! 🌟
This recognition is specifically aimed at software and solution partners leveraging NVIDIA technology in the cloud to provide quick and effective customer outcomes.
NetsPresso has also been included in the prestigious NVIDIA AI Accelerated ecosystem, which only features the most exceptional solutions recognized by NVIDIA. This recognition from renowned global companies further emphasizes the importance and excellence of NetsPresso in the edge AI space. 🏆
|
|
⌨️
NetsPresso® and the New Lineup
|
|
|
Nota AI has recently released two new products based on the AI model optimization technology incorporated in NetsPresso®.
The first product, PyNetsPresso, allows developers to create AI models using Python code. The second product is LaunchX, streamlines and simplifies the process of deploying AI models to devices. Users can access both products seamlessly with a single NetsPresso account. 👍🏻
The following sections provide a detailed introduction to each of these products. 🤭
|
|
NetsPresso® is a hardware-aware AI model optimization platform. It reduces the time and resources required to develop an AI model and optimizes the model for target hardware to enable a more efficient and effective deployment in production.
NetsPresso consists of three modules: Model Searcher, Model Compressor, and Model Launcher. These modules work together to automatically search, compress, and deploy optimized AI models for target hardware. AI engineers have the flexibility to customize the modules that best suit their development stages and goals, resulting in highly optimized AI models.
Create optimized AI models for target devices through NetsPresso! 🙂
|
|
PyNetsPresso, a Python-based library, has been introduced to replace the previous web-based service of NetsPresso®. 🚀
This new release allows you to develop your models into optimized AI models using the code repositories on GitHub. With PyNetsPresso, you can easily create optimized AI models using Python code.
Streamline your AI model optimization workflow with PyNetsPresso!
|
|
Introducing LaunchX, a standalone service that offers the Model Launcher feature previously available on NetsPresso®️. LaunchX allows users to easily upload their AI models and convert them efficiently in just two simple steps. This innovative tool enables the generation of benchmarks on different devices, helping users identify the best device for optimal performance. Moreover, LaunchX provides free models that can be customized to suit specific devices.
Give LaunchX a try today and upload models optimized for your target devices! 🤗
|
|
On May 4, members from Nota AI’s NetsPresso Application team, including Hyungjun Lee, Wooksu Shin, and Hancheol Park, achieved an amazing result in The VOTS2023 Challenge held within the ICCV 2023 workshop. The MCMOT system we presented at the Challenge was ranked in the top 10 globally for Accuracy and Quality! 🎉
With this achievement, Nota AI was selected as a co-author of the results paper for the Challenge. Let's learn more about both of these news together.
|
|
|
Ranked #2 in Accuracy & #8 in Quality
|
|
The MCMOT system submitted by Nota AI's NetsPresso Application team achieved 2nd place in the Accuracy category and 8th place in the Quality category. The high performance of the system, designed with MixFormer and SAM methodology, was recognized in both categories.
The Challenge brought together 47 leading teams with tracking technology on a global scale, making it even more meaningful for Nota AI's global presence in the field of tracking technology. 🌍
|
|
|
Selected as a Co-author of the Results Paper
|
|
|
As a result of our excellent results in the accuracy and quality categories, we were honored to be selected as a co-author of the VOTS2023 results paper. This award is given to the team that submits a system with noteworthy performance among the teams participating in the Challenge and is a great recognition of Nota AI's technology. 🔥
|
|
🎉
Received KCC 2023 Best Paper Award
in the AI field
|
|
|
The joint research paper by members from Nota AI’s NetsPresso Application team, including Donghyuk Choi, Wooksu Shin, and Hancheol Park, along with members from Hanyang University and Soongsil University, has been honored with the Best Paper Award at the Korean Computer Conference (KCC) hosted by the Korean Institute of Information Scientists and Engineers!
The winning paper, "Lightweight Semantic Segmentation Model Compression using Weighted Sum-based Feature Fusion Decoder," explores the tradeoff between the performance and efficiency of AI models and how effective decoding methods can be used to achieve good semantic segmentation results with fewer resources.
As a result of winning the Best Paper Award for Artificial Intelligence, we have been invited to submit a paper to the journal, and our paper will be published. This is especially meaningful because it was held at a conference organized by the Korea Information Science Society, the oldest and largest organization in the field of computer science in Korea.
|
|
📃
Articles Published on AI Model Optimization using NetsPresso®
|
|
|
NetsPresso®, a hardware-aware AI model optimization platform, applies to various industries, allowing for the easy creation of lightweight AI models.
This section introduces two articles that applications developed using the optimized AI models generated with NetsPresso.
|
|
|
Advancing Road Safety with AI-Powered Pothole Detection
|
|
Pothole Detection, the app utilizes a lightweight AI model created with NetsPresso®. This advanced application enables real-time detection of potholes on the road, instantly alerting drivers and ensuring their safety.
Our app utilizes an optimized AI model generated with NetsPresso, focusing on identifying potholes. By leveraging NetsPresso, we have achieved outstanding accuracy in detecting potholes while keeping latency to a minimum.
To learn more, click the button below and check out the article uploaded in the Community tab on Nota AI's official website.
|
|
|
Laundry Symbol Detection with AI Model Optimization
|
|
Introducing Nota AI's laundry symbol detection application powered by NetsPresso AI model optimization! 🔊 This application provides the ability to detect and interpret laundry symbols through AI model optimization. Since the AI model is optimized through NetsPresso, the results are highly accurate and simplified.
Normally, laundry symbols are difficult to interpret because they are very small in size and have a complex system, but this application makes it easy. If you're tired of deciphering cryptic laundry symbols or curious about AI's potential in optimizing everyday processes, this article is a must-read!
|
|
|
|
