The 15th CSIG Enterprise Visit at Qihoo 360
On June 29, 2023, the "CSIG Corporate Tour - Enter Qihoo 360" hosted by the Chinese Society of Image and Graphics (CSIG) and hosted by Beijing Qihoo Technology Co., Ltd. and the CSIG Youth Working Committee was successfully held, attracting more than 1,000 people online and offline to participate in the event.
The theme of this event is "multimodal and cross-modal Learning in the Big Model Era". Experts and scholars from Peking University, Zhejiang University, Harbin Institute of Technology, Wuhan University, as well as teachers and students from universities were specially invited to discuss the cutting-edge trends in the industry with the Qihoo 360 technical team.
At the beginning of the event, Liu Si, deputy secretary-general of the Chinese Society of Graphics and Graphics and professor at Beihang University, gave an online speech, detailing the society’s development history, organizational structure, academic activities, industry-university-research integration, talent recommendation and other work. He also expressed his sincere gratitude to the guest speakers for bringing cutting-edge views and technologies in the field, as well as Qihoo 360’s strong support for the event.
Subsequently, Yin Yuhui, vice president of 360 Group, head of 360 Technology Center, and chairman of the professional committee, delivered a welcome speech and expressed a warm welcome to the society and all guests. He said that large multimodal model (LMM) is currently the hottest research focus in the industry and academia, and is also the only way forward for general artificial intelligence. Teachers and students are witnessing and actually participating in the transformative moment of the development of artificial intelligence. Large models are turning "oil" such as data and knowledge into "water and electricity" in the intelligent era, and are linking it to our production and life step by step. The revolution in related fields has just begun. Yin Yuhui further shared 360 Group's artificial intelligence development strategy of "focusing on core capability building with one hand and application scenarios with the other", as well as the latest production and research progress of "Intelligent Brain 4.0".
Figure 2 Vice President Yin Yuhui’s speech
In the academic report session, experts and scholars from universities and scientific research institutes shared the results of their respective research fields with everyone.
Professor Zhao Zhou of Zhejiang University made "cross-modal Research on Audio and Video Generation Model" Theme report shares the real-time, high-quality, lightweight, and generalizable speech synthesis NATSpeech work for multimodal human-computer interaction scenarios; the high-performance, multi-task, and transferable singing voice synthesis DiffSinger work; the open, sequential, multi-task AudioGPT work, and the controllable generalization, efficient, robust, and modally versatile face video synthesis GeneFace work.
Figure 3 Professor Zhao Zhou gives a report
Professor Wu Yu from Wuhan University gave a report on "multimodal Perception and Generation", sharing multimodal learning including audio and video understanding, visual-language and other perceptual models, as well as a series of the latest multimodal generation methods based on diffusion model.
Figure 4 Professor Wu Yu gives a report
Zhang Zheng, associate professor of Harbin Institute of Technology (Shenzhen), gave a report on "Multi-Source Interactive Emotion Understanding and Analysis", focusing on the latest research progress in face, voice and multimodal emotion understanding and analysis, involving audio-visual emotional feature extraction, cross-domain emotion analysis and multimodal algorithms and related applications.
Figure 5 Associate Professor Zhang Zheng giving a report
He Xiangteng, an assistant researcher at the Wangxuan Institute of Computer Science at Peking University, gave a keynote report on "fine-grained Cross-media Classification and Retrieval", detailing the research status and progress of fine-grained cross-media classification and retrieval, and discussing future research directions.
Figure 6 Assistant Researcher He Xiangteng gives a report
Dr. Leng Dawei, head of the visual engine of 360 Artificial Intelligence Research Institute, presented "multimodal and cross-modal learning in the era of large models" The theme report summarizes the recent work progress in the MLLM direction from the perspective of the industry. It analyzes the two current research routes: the native multimodal route and the single-modal expert model stitching route. It also introduces the 360 Artificial Intelligence Research Institute’s research and development thinking, recent results, and future work directions in the MLLM direction.
Figure 7 Dr. Leng Dawei gives a report
During the event, under the leadership of the technical center, the guests visited Qihoo 360's exhibition hall and learned in detail about Qihoo 360's development history, special strategic positioning and special contribution to the national cyberspace security defense line. They felt first-hand Qihoo 360's unremitting efforts in digital security, digitization, and artificial intelligence, as well as its outstanding achievements and honors achieved by relying on top technology.
Figure 8 Visit and exchange
This CSIG Enterprise Tour event was a complete success, effectively promoting the interaction between universities, research institutes and enterprises. In the next step, the society will continue to promote the integrated development of industry, academia, and research, and actively build a technology and talent docking platform for enterprises. It also hopes to strengthen cooperation with more enterprises in the field and work together to promote the development of the image and graphics field.