Research

AAAI 2025 | Interior Spatial Modal Awareness for 3D Anomaly Detection

AAAI 2025 | Two-Stage Healthcare Reporting Framework (DAMPER)

Professor Shen Linlin's team won second place in the National Intelligent Agent Design Competition for their collaborative multi-agent approach to targeted environmental protection education for citizens.

Good news! The School of Artificial Intelligence has secured 15 National Natural Science Foundation of China projects in its first batch of projects for 2025, setting a new record!

FROGENT: The first end-to-end intelligent agent for drug design.

Research on key theories and technologies of multimodal collaborative perception in the Internet of Things

This paper explores multimodal collaborative sensing in the Internet of Things (IoT), proposes a theoretical and methodological system for wireless sensing in IoT based on multimodal sensing, constructs an efficient collaborative method and network computing decision model, and proposes a collaborative security theory and method, comprehensively improving the sensing capability, collaborative capability and security assurance capability of IoT systems. (1) Multimodal sensing: Combining WiFi and RFID equal-mode signals, the application boundaries of wireless sensing are expanded, realizing efficient human and object sensing in IoT based on wireless. (2)

Intelligent medical image analysis

The Visual Big Data Research Center is developing the "Lingli Medical" medical big data model, which generates diagnostic reports based on images. It is also developing intelligent analysis algorithms for medical images, including OCT images of the eyes, nasal endoscopy images, digital oral and orthodontic images, ultrasound images of thyroid tumors, chest X-ray images, gastrointestinal endoscopy images, kidney CT images, and histopathological and cytopathological images. Furthermore, it is combining clinical diagnostic and laboratory data to conduct research on computer-aided precision medicine methods for multimodal disease diagnosis and prognosis.

Intelligent analysis of industrial big data

The Industrial Big Data Intelligent Analysis Project developed the world's first intelligent robot for detecting surface defects in the containment vessel of nuclear power plants, effectively ensuring the safety of critical facilities in my country. The robot achieves a crack width recognition error of 0.05mm, a defect location error of less than 10cm, and a defect detection rate greater than 90%. Key technologies include: 1. A SLAM algorithm that fuses point and line features, extracting corner points from consecutive frames, using KLT optical flow to track these corner points, and employing an LSD line segment detector to extract line features. 2. A multi-branch detection model based on a rotating frame crack detection perspective, using a loss function based on geometric distance perception to extract differential depth features. This technology...

Lingli Chinese big language model

Lingli Chinese Large Language Model, based on TencentPreTrain (the world's first modular large-scale model pre-training framework with 100,000 downloads per month, published at the top natural language processing conference ACL 2023), is the first open-source Chinese 7B, 13B, and 30B large-scale model. Its core technologies include: (1) LLaMA Chinese pre-training, instruction fine-tuning, and question answering; (2) language transfer learning through adaptive data sampling; (3) English → parallel Chinese-English corpus training; and (4) adaptive adjustment of training ratio to address knowledge forgetting and transfer.

Key technologies for cloud data exchange and service equipment

The research on key technologies and industrialization of cloud data exchange and service equipment focuses on three fundamental aspects: cloud security, cloud interaction, and cloud performance. It conducts theoretical research and technological breakthroughs in addressing diverse cloud security needs, massive cloud resource allocation and scheduling, and the security of large-scale cloud data processing and exchange. This achievement won the first prize of Guangdong Provincial Science and Technology Award and the first prize of the Chinese Institute of Electronics Science and Technology Award.