Gathering with "Sounds" | RTE 2021 Programming Challenge Concludes Successfully


"RTE 2021 Programming Challenge" co-sponsored by Agora and Huanxin came to a successful conclusion. The 46 finalists that broke out of 200+ participating teams used their wonderful defenses to draw a successful conclusion to the contest that lasted for more than two months.

This year’s "RTE 2021 Innovative Programming Challenge" is divided into 2 tracks: application innovation track continues the "use the sound network Agora SDK to develop applications" question; technology innovation track developers can "use Soundnet cloud market plug-in interface, develop self-developed plug-ins and function demonstration Demo".

Although the competition system and questions put forward higher requirements for the participating teams, it also provides unique technological innovation space for everyone. Compared with last year, two tracks are almost twice as much as last year.

The finals and awards of this competition were all conducted online through Agora Video Call App, and the whole process was broadcast live through station B. In the end, the finals gave birth to a team with the first, second and third prizes of the application innovation track, one for the "Huanxin Special Award" and six for the "Excellence Award"; the technological innovation track for the "Technology Innovation Special Award" One and one "Excellence Award".


Application innovation track

First prize: Agora Home AI

With the improvement of the performance of smart devices and the rapid development of networks, smart hardware based on audio and video is also booming. Cross-brand and cross-product device management has also become a topic that lingers in the daily use of users.

The "Agora Home AI" system is based on the theme of smart home, using cloud signaling SDK to realize remote control of IoT devices. At the same time, the 1V1 video of human and machine is realized through the Agora RTC SDK of the sound network, and the video collected by the robot is sent to the PC control center for AI intelligent detection, and event response is triggered.


The system uses the open source Yolo V3 algorithm to process various video data, supports C#, C++ calls; Unity 3D, VS series development. At present, it has supported Yolo's basic 80 object recognition, helmet recognition, and ice hockey recognition files. Use the cloud signaling SDK provided by Soundnet for remote device control, build group rooms for real-time message communication, and support intelligent hardware control through custom protocols.

"Agora Home AI" can help users realize how to control the access of wearable devices, smart furniture devices, and video surveillance devices. Including smart lighting, smart doors and windows, smart door locks, smart security, smart bracelet monitoring, smart home appliance control and other supporting products, allows users to realize the interconnection, unified management, and smart linkage of smart devices of multiple brands in a unified interactive platform . To create a more comfortable, safer and more energy-saving home living environment for users.

Second prize: Agora FIow

The second-placed work "Agora Flow" is a audio and video Low Code Web sharing editor based on the sound network + Huanxin SDK.

The inspiration for the work comes from the use of the Agora SDK. In the process of using the Agora SDK, creators have been thinking about audio and video services in addition to providing services in the form of SDK and empowering developers, is there any other form? Low Code is such a possible solution. The audio and video related functions are modularized and integrated, and a graphical interface is provided for developers to use in the form of PPT to complete the functions they want to achieve.

The work provides a web-based template project integrated with RTC Chat SDK through the audio and video transmission of the sound network and cloud signaling SDK products, and the configuration items are generated through CodeGen. Realized the automatic generation of the Low Code project of the online flowchart editor. can be done by dragging and dropping 160ee41b1b6880.


Some developers joked that many of the works in this competition are developing an App, and "Agora Flow" is a project to help developers better develop App.

Third Prize: Urban Explorer

The "Urban Explorer" project is a social product built using map LBS + cloud signaling SDK + real-time audio and video. This product is designed for friends who want to find new friends with common hobbies and travel and explore together in the city.


The use of the product is very simple and convenient. After the user has registered and logged in, the location of the user will be automatically updated through the LBS map. Just click "initiate a task" and select the number of "expeditions", and the user can have a brand new urban adventure with their friends.

For the RTE scenario, a multi-person chat room scenario is implemented in the product. In addition, 160ee41b1b6908 combines actual business scenarios through the use of cloud signaling SDK product will also be connected to the video chat function in the future, so that small partners who have no way to travel in real time can also participate in the city's adventure.

Huanxin Special Award: Wangyou Hall

"Forget Worry Museum" is a very warm work, which hopes to help people in modern life forget their worries and relieve their sorrows by talking to each other, and spread positive energy.

This is a social app for strangers, including information flow. It combines several of the most common forms of social products, including text chat, phone calls, and so on. Let some troubles that are inconvenient to talk to relatives and friends, you can find sympathy and comfort when communicating with strangers.

Excellence Award: Vchat

"Vchat" uses facial bone recognition and cloud signaling SDK to realize real-time calls with virtual 3D characters. Use the WebGL engine of tensorflow.js as the backend, and use the existing open source face recognition model to recognize the position of the face and the state of the facial features through the camera. Then update the face data to the virtual 3D model in real time through Three.js and Vrm.js.


In the video part, the face skeleton data is transmitted to the channel in real time through the real-time message RTM SDK so that other users can subscribe to restore the face. The voice part transmits the sound in real time through the RTC SDK and allows users to subscribe. It can realize the functions of synchronous face changing and voice changing chat.

In addition to the aforementioned "Vchat", there are "Smart Class Answering Components", "Agora X-Runtime", "Weln", "Huanxin (", "Smart AR Brush Copy Teaching System/CopyTeachWorks" The Excellence Award in the "Application Innovation Track" of this competition. There may be no way to present these excellent works to everyone here. Interested friends can go to our Github repository to view:

Technological innovation track

Special Award for Technological Innovation: Face Recognition

The "Technological Innovation Special Award" is an award specially set up for the "Technological Innovation Track". The winning work is a "Face Recognition" plug-in based on C++ language package developed using AgoraEngineKit2 on the iOS platform.

The work interacts with the outside through the TYSMExtensionManger class, and internally deals with the relevant logic implemented by the plug-in. Put IExtensionProvider, IVideoFilter and your own developed Processor in the same place. The two files are publicly disclosed in framework mode. is not only convenient for developers to read, but also can be passed as a parameter to enhance code readability .


The plug-in can support face detection, tracking, and tracking and recognition of multiple faces, and provide feedback in 3D points on the recognized areas such as facial contours, eyes, eyebrows, nose, and mouth.

Excellence Award: Water Mask

The "Water Mask" project is an entry in the "Technology Track". Through the video capture or playback link of the Shengwang SDK, an invisible watermark of image or text type is added to the YUV domain or after encoding.


After the hidden watermark (blind watermark) is added, the user cannot directly see the watermark information in the video. While protecting the copyright of the video publisher, it also guarantees the user's video viewing experience. In the future, "Water Mask" also hopes to expand the voiceprint watermark in audio processing, bringing more and better experiences to the industry in scenarios such as .

The above are some of the winning works and team situation of this "RTE 2021 Programming Challenge". More works about this challenge will be open sourced on Github, and interested friends can check it out:

In addition to the prize money of this quarter, the winning team can also apply to enter the Agora recruitment green channel and enjoy the benefits of the entrepreneurial support plan within one year. For details, please click here [Official website] to view.


阅读 8.1k


SegmentFault 思否编辑,欢迎投稿优质技术资讯!

871 声望
217 粉丝
0 条评论

SegmentFault 思否编辑,欢迎投稿优质技术资讯!

871 声望
217 粉丝