Search

Your search keyword '"Shen, Qianqian"' showing total 516 results
Start Over Author "Shen, Qianqian"
516 results on '"Shen, Qianqian"'

1. SAM-SP: Self-Prompting Makes SAM Great Again

Author

Zhou, Chunpeng, Ning, Kangjie, Shen, Qianqian, Zhou, Sheng, Yu, Zhi, and Wang, Haishuai

Subjects
Computer Science - Computer Vision and Pattern Recognition, Computer Science - Artificial Intelligence, Computer Science - Emerging Technologies
Abstract

The recently introduced Segment Anything Model (SAM), a Visual Foundation Model (VFM), has demonstrated impressive capabilities in zero-shot segmentation tasks across diverse natural image datasets. Despite its success, SAM encounters noticeably performance degradation when applied to specific domains, such as medical images. Current efforts to address this issue have involved fine-tuning strategies, intended to bolster the generalizability of the vanilla SAM. However, these approaches still predominantly necessitate the utilization of domain specific expert-level prompts during the evaluation phase, which severely constrains the model's practicality. To overcome this limitation, we introduce a novel self-prompting based fine-tuning approach, called SAM-SP, tailored for extending the vanilla SAM model. Specifically, SAM-SP leverages the output from the previous iteration of the model itself as prompts to guide subsequent iteration of the model. This self-prompting module endeavors to learn how to generate useful prompts autonomously and alleviates the dependence on expert prompts during the evaluation phase, significantly broadening SAM's applicability. Additionally, we integrate a self-distillation module to enhance the self-prompting process further. Extensive experiments across various domain specific datasets validate the effectiveness of the proposed SAM-SP. Our SAM-SP not only alleviates the reliance on expert prompts but also exhibits superior segmentation performance comparing to the state-of-the-art task-specific segmentation approaches, the vanilla SAM, and SAM-based approaches., Comment: Under Review

Published
2024

5. A High-Resolution Dataset for Instance Detection with Multi-View Instance Capture

Author

Shen, Qianqian, Zhao, Yunhan, Kwon, Nahyun, Kim, Jeeeun, Li, Yanan, and Kong, Shu

Subjects
Computer Science - Computer Vision and Pattern Recognition
Abstract

Instance detection (InsDet) is a long-lasting problem in robotics and computer vision, aiming to detect object instances (predefined by some visual examples) in a cluttered scene. Despite its practical significance, its advancement is overshadowed by Object Detection, which aims to detect objects belonging to some predefined classes. One major reason is that current InsDet datasets are too small in scale by today's standards. For example, the popular InsDet dataset GMU (published in 2016) has only 23 instances, far less than COCO (80 classes), a well-known object detection dataset published in 2014. We are motivated to introduce a new InsDet dataset and protocol. First, we define a realistic setup for InsDet: training data consists of multi-view instance captures, along with diverse scene images allowing synthesizing training images by pasting instance images on them with free box annotations. Second, we release a real-world database, which contains multi-view capture of 100 object instances, and high-resolution (6k x 8k) testing images. Third, we extensively study baseline methods for InsDet on our dataset, analyze their performance and suggest future work. Somewhat surprisingly, using the off-the-shelf class-agnostic segmentation model (Segment Anything Model, SAM) and the self-supervised feature representation DINOv2 performs the best, achieving >10 AP better than end-to-end trained InsDet models that repurpose object detectors (e.g., FasterRCNN and RetinaNet)., Comment: Accepted by NeurIPS 2023, Datasets and Benchmarks Track

Published
2023

6. Q-Net: Query-Informed Few-Shot Medical Image Segmentation

Author

Shen, Qianqian, Li, Yanan, Jin, Jiyong, and Liu, Bin

Subjects
Computer Science - Computer Vision and Pattern Recognition
Abstract

Deep learning has achieved tremendous success in computer vision, while medical image segmentation (MIS) remains a challenge, due to the scarcity of data annotations. Meta-learning techniques for few-shot segmentation (Meta-FSS) have been widely used to tackle this challenge, while they neglect possible distribution shifts between the query image and the support set. In contrast, an experienced clinician can perceive and address such shifts by borrowing information from the query image, then fine-tune or calibrate her prior cognitive model accordingly. Inspired by this, we propose Q-Net, a Query-informed Meta-FSS approach, which mimics in spirit the learning mechanism of an expert clinician. We build Q-Net based on ADNet, a recently proposed anomaly detection-inspired method. Specifically, we add two query-informed computation modules into ADNet, namely a query-informed threshold adaptation module and a query-informed prototype refinement module. Combining them with a dual-path extension of the feature extraction module, Q-Net achieves state-of-the-art performance on widely used abdominal and cardiac magnetic resonance (MR) image datasets. Our work sheds light on a novel way to improve Meta-FSS techniques by leveraging query information., Comment: Accpeted by Intelligent Systems Conference (IntelliSys) 2023

Published
2022

7. Q-Net: Query-Informed Few-Shot Medical Image Segmentation

Author

Shen, Qianqian, Li, Yanan, Jin, Jiyong, Liu, Bin, Kacprzyk, Janusz, Series Editor, Gomide, Fernando, Advisory Editor, Kaynak, Okyay, Advisory Editor, Liu, Derong, Advisory Editor, Pedrycz, Witold, Advisory Editor, Polycarpou, Marios M., Advisory Editor, Rudas, Imre J., Advisory Editor, Wang, Jun, Advisory Editor, and Arai, Kohei, editor

Published
2024
Full Text
View/download PDF

15. Volumetric prostate segmentation : from multi atlas method to deep learning models

Author

Shen, Qianqian, Laurenson, David, and Nailon, William Henry

Abstract

Accurately segmenting the prostate gland in magnetic resonance (MR) images provides a valuable assessment of the prostate volume for diseases diagnosis as well as treatment planning. In clinical practice, prostate segmentation is predominantly undertaken by experienced clinical oncologists based on slice-by-slice visual inspection, which is expensive, time-consuming, and prone to inter-observer and intra-observer variability. Therefore, it is significantly crucial and essential to ensure techniques for automatic, accurate and efficient segmentation of the prostate gland are consistently developed. The focus of this thesis is on the development of volumetric segmentation techniques in MR images of the prostate utilising multi-atlas method, as well as deep learning models. Firstly, a feature-specific atlas pre-selection is proposed to maximise computation efficiency through the utilisation of statistical texture features as handcrafted features and latent features learned from an autoencoder (AE). The proposed technique could cluster atlases into groups without pre-alignment and accomplish a mean dice similarity coefficient (DSC) of 0.852 ± 0.028 with texture features learned from the region of the prostate and its surrounding tissues. Secondly, in the present study, a 3D convolutional neural network (CNN) is developed through the introduction of separable and anisotropic convolutions for reducing the number of parameters that can be trained. The 3D model with light parameters proposed in this study is capable of producing more precise and spatially continuous prostate boundaries with a mean DSC value of 0.883 ± 0.031 and outperform V-Net with 97% and 96% reduction in parameters and floating-point operations (FLOPs). Thirdly, in the current study, using thin-3D volume with fewer slices as inputs is proposed for reducing computation cost as well as preserving contextual information along the z-axis for volumetric segmentation. The thin-3D model with the thickness of 13 has only 3% of V-Net's parameters and 10% of its FLOPs compared to V-Net, nevertheless, the thin-3D model can achieve a mean DSC of 0.908 ± 0.012. Lastly, to ensure the enhancement in the efficiency of training, the large-scaled unlabelled data from the target domain is leveraged with small-scaled labelled data. Regarding the prostate segmentation task, the pre-training of the convolutional autoencoder (CAE) is conducted for the image denoising task on unlabelled data and then refined on the labelled data. This method proposed is capable of generating a mean DSC of 0.893 ± 0.023, which when compared to the baseline is 2.3% better. Based on the DSC scores, the atlas-based method performs worse than the other three deep learning methods. Separable and anisotropic convolutions and thin-3D volume could assist in reducing the model complexity and computation cost. It is feasible for the pre-training strategy that applies unlabelled target domain data to learn feature representations that correspond with labelled data, enhances training efficiency during fine-tuning, as well as decreases the effect of overfitting.

Published
2022
Full Text
View/download PDF

Catalog

Books, media, physical & digital resources
See catalog results