A5082831514- Advanced Neural Network Applications
- Advanced Image and Video Retrieval Techniques
- Robotics and Sensor-Based Localization
- Remote Sensing and LiDAR Applications
- Advanced Memory and Neural Computing
- Advancements in Battery Materials
- Image Retrieval and Classification Techniques
- Remote Sensing and Land Use
- Multimodal Machine Learning Applications
- Industrial Vision Systems and Defect Detection
- Face recognition and analysis
- AI in cancer detection
- Domain Adaptation and Few-Shot Learning
- Advanced Battery Technologies Research
- QR Code Applications and Technologies
- Supercapacitor Materials and Fabrication
- CCD and CMOS Imaging Sensors
- Facial Nerve Paralysis Treatment and Research
- Visual Attention and Saliency Detection
- Medical Image Segmentation Techniques
Shanghai Artificial Intelligence Laboratory
2023
ShangHai JiAi Genetics & IVF Institute
2023
Dalian University of Technology
2022
Beijing University of Chemical Technology
2020-2021
Chongqing University of Posts and Telecommunications
2019
Compared to the great progress of large-scale vision transformers (ViTs) in recent years, models based on convolutional neural networks (CNNs) are still an early state. This work presents a new CNN-based foundation model, termed InternImage, which can obtain gain from increasing parameters and training data like ViTs. Different CNNs that focus large dense kernels, InternImage takes deformable convolution as core operator, so our model not only has effective receptive field required for...
We present the All-Seeing (AS) project: a large-scale data and model for recognizing understanding everything in open world. Using scalable engine that incorporates human feedback efficient models loop, we create new dataset (AS-1B) with over 1 billion regions annotated semantic tags, question-answering pairs, detailed captions. It covers wide range of 3.5 million common rare concepts real world, has 132.2 tokens describe their attributes. Leveraging this dataset, develop (ASM), unified...
Object detection and instance segmentation in remote sensing images is a fundamental challenging task, due to the complexity of scenes targets. The latest methods tried take into account both efficiency accuracy segmentation. In order improve them, this paper, we propose single-shot convolutional neural network structure, which conceptually simple straightforward, meanwhile makes up for problem low networks. Our method, termed with SSS-Net, detects targets based on location object's center...
Object detection and semantic segmentation have achieved remarkable performance propelled by deep convolutional neural networks. However, neither of them can well parse deal with swarms rotating ships in remote sensing images. In this article, we pay more attention to the instance-level task, which recognizes objects effectively straightly. We propose a new network architecture, called orientated silhouette matching network, employing multiscale features masks enable single-shot...
Compared to the great progress of large-scale vision transformers (ViTs) in recent years, models based on convolutional neural networks (CNNs) are still an early state. This work presents a new CNN-based foundation model, termed InternImage, which can obtain gain from increasing parameters and training data like ViTs. Different CNNs that focus large dense kernels, InternImage takes deformable convolution as core operator, so our model not only has effective receptive field required for...
In this paper, we introduce a brand new dataset to promote the study of instance segmentation for objects with irregular shapes. Our key observation is that though irregularly shaped widely exist in daily life and industrial scenarios, they received little attention field due lack corresponding datasets. To fill gap, propose iShape, an shape segmentation. iShape contains six sub-datasets one real five synthetics, each represents scene typical shape. Unlike most existing datasets regular...
The rapid progress of photorealistic synthesis techniques has reached a critical point where the boundary between real and manipulated images starts to blur. Recently, mega-scale deep face forgery dataset, ForgeryNet which comprised 2.9 million 221,247 videos been released. It is by far largest publicly available in terms data-scale, manipulations (7 image-level approaches, 8 video-level approaches), perturbations (36 independent more mixed perturbations), annotations (6.3 classification...
In order to meet the requirements of image retrieval and real-time index storage massive datasets, a Content-based Image Retrieval (CBIR) based on vector quantization is used in this paper. By feature extraction depth residual network, improved Product Quantization (PQ) coding algorithm generate dynamic database under FAISS Framework, which can realize nonlinear large-scale CBIR with high robustness. Empirical results show that PQ has achieved good performance compression rate, recall rate...
Object detection and instance segmentation in remote sensing images is a fundamental challenging task, due to the complexity of scenes targets. The latest methods tried take into account both efficiency accuracy segmentation. In order improve them, this paper, we propose single-shot convolutional neural network structure, which conceptually simple straightforward, meanwhile makes up for problem low networks. Our method, termed with SSS-Net, detects targets based on location object's center...
With the development of technology and automation modern industry, finding a new kind energy storing device gradually occupies significant position. The supercapacitor takes up scientists’ eyes, because its advantages high watt density, long servicing life non-pollution. However, there is also some barriers blocking way forward development. In this project, we use MOF-5 as electrode material, combining with other metal atoms to enhance specific capacitance. Through researching, found that...