This page documents the open source projects created by, or receive significant contributions from, lab members.
Contact: yanye.lu@pku.edu.cn
| Index | Name | Infomation | Introduction | Access |
|---|---|---|---|---|
| OSP-21 | Choroidal Optical Coherence Tomography Angiography: Non-invasive Choroidal Vessel Analysis via Deep Learning | Date Created Owner Implement in | TBD | Repository Data License Related Paper |
| OSP-20 | DA-WSSS: Boosting Weakly Supervised Object Localization and Segmentation with Domain Adaption
| Date Created Owner Implement in | The DA-WSOL pipeline is elaborated to better assist WSOL with DA approaches by considering the specificities for the adaption of WSOL. Our DA-WSOL pipeline can discern the source-related and the Universum samples from other target samples based on a proposed target sampling strategy and then utilize them to solve the sample unbalancing and label unmatching between the source and target domain of WSOL. Experiments show that our pipeline outperforms SOTA methods on three WSOL benchmarks and can improve the performance of downstream weakly supervised semantic segmentation tasks. The related work has been published by IEEE Transactions on Pattern Analysis and Machine Intelligence. | Repository License Related Paper |
| OSP-19 | V2L-Tokenizer: Beyond Text: Frozen Large Language Models in Visual Signal Comprehension
| Date Created Owner Implement in | We present the Vision-to-Language Tokenizer, abbreviated as V2T Tokenizer, which transforms an image into a ``foreign language'' with the combined aid of an encoder-decoder, the LLM vocabulary, and a CLIP model. With this innovative image encoding, the LLM gains the ability not only for visual comprehension but also for image denoising and restoration in an auto-regressive fashion-crucially, without any fine-tuning. This work has been accepted by CVPR2024. | Repository License Related Paper |
| OSP-18 | PCNet: Prior Category Network for CT Universal Segmentation Model
| Date Created Owner Implement in | Inspired by how doctors identify tissues, we propose a novel approach, the Prior Category Network (PCNet), that boosts segmentation performance by leveraging prior knowledge between different categories of anatomical structures. Our PCNet comprises three key components: prior category prompt (PCP), hierarchy category system (HCS), and hierarchy category loss (HCL). The related work has been published by IEEE Transactions on Medical Imaging. | Repository License Related Paper |
| OSP-17 | BraSEDA: Structure-Enhanced Unsupervised Domain Adaptation for CT Whole-Brain Segmentation
| Date Created Owner Implement in | We propose BraSEDA, a CT-based unsupervised domain adaptation (UDA) model designed to assist in the identification of brain regions. BraSEDA framework utilizes a cross-modal instance normalization (CMIN) module for enhancing CT image structural features and creating high-quality pseudo MR images. A multi-level CMIN architecture is also introduced for further improvement. The related work has been published by IEEE Transactions on Radiation and Plasma Medical Sciences. | Repository License Related Paper |
| OSP-16 | PSCAT: a lightweight transformer for simultaneous denoising and super-resolution of OCT images
| Date Created Owner Implement in | In this paper, we propose a lightweight Transformer to efficiently reconstruct high-quality images from noisy and low-resolution OCT images acquired by short scans. Our method, PSCAT, parallelly employs spatial window self-attention and channel attention in the Transformer block to aggregate features from both spatial and channel dimensions. It explores the potential of the Transformer in denoising and super-resolution for OCT, reducing computational costs and enhancing the speed of image processing. The related work has been published by Biomedical Optics Express. | Repository License Related Paper |
| OSP-15 | CDSP: Scribble Hides Class: Promoting Scribble-Based Weakly-Supervised Semantic Segmentation with Its Class Label
| Date Created Owner Implement in | In this study, we propose a class-driven scribble promotion network, which utilizes both scribble annotations and pseudo-labels informed by image-level classes and global semantics for supervision. Directly adopting pseudo-labels might misguide the segmentation model, thus we design a localization rectification module to correct foreground representations in the feature space. This work has been accepted by AAAI2024. | Repository License Related Paper |
| OSP-14 | RD3D: 3-D Convolutional Neural Networks for RGB-D Salient Object Detection and Beyond
| Date Created Owner Implement in | We propose to disentangle the conventional 3-D convolution into successive spatial and temporal convolutions and, meanwhile, discard unnecessary zero padding. This eventually results in a 2-D convolutional equivalence that facilitates optimization and reduces parameters and computation costs. Thanks to such a progressive-fusion strategy involving both the encoder and the decoder, effective and thorough interactions between the two modalities can be exploited and boost detection accuracy. The related work has been published by IEEE Transactions on Neural Networks and Learning Systems. | Repository License Related Paper |
| OSP-13 | One-Pot Multi-Frame Denoising
| Date Created Owner Implement in | In this paper, we propose a novel unsupervised learning strategy called one-pot denoising (OPD), which is the first unsupervised multi-frame denoising method. We provide two specific implementations, namely OPD-random coupling and OPD-alienation loss, to achieve OPD during model training based on data allocation and loss refine, respectively. The related work has been published by International Journal of Computer Vision. | Repository License Related Paper |
| OSP-12 | BCAM: Background-Aware Classification Activation Map for Weakly Supervised Object Localization
| Date Created Owner Implement in | Our work proposes a novel mechanism called the background-aware classification activation map (B-CAM) to add background awareness for WSOL training. Besides aggregating an object image-level feature for supervision, our B-CAM produces an additional background image-level feature to represent the pure-background sample. This additional feature can provide background cues for the object classifier to suppress the background activations on object localization maps. Moreover, our B-CAM also trained a background classifier with image-level annotation to produce adaptive background scores when determining the binary localization mask. The related work has been published by IEEE Transactions on Pattern Analysis and Machine Intelligence. | Repository License Related Paper |
| OSP-11 | DEMCR: Discriminative ensemble meta-learning with co-regularization for rare fundus diseases diagnosis
| Date Created Owner Implement in | In this work, we develop an automatic diagnosis system for rare fundus diseases, based on the meta-learning framework. The system incorporates a co-regularization loss and the ensemble-learning strategy into the meta-learning framework, fully leveraging the advantage of multi-scale hierarchical feature embedding. We initially conduct comparative experiments on our newly-constructed lightweight multi-disease fundus images dataset for the few-shot recognition task (namely, FundusData-FS). The related work has been published by Medical Image Analysis. | Repository Data License Related Paper |
| OSP-10 | BagCAMs: Bagging Regional Classification Activation Maps for Weakly Supervised Object Localization
| Date Created Owner Implement in | This paper elaborates a plug-and-play mechanism called BagCAMs to better project a well-trained classifier for the localization task without refining or re-training the baseline structure. Our BagCAMs adopts a proposed regional localizer generation (RLG) strategy to define a set of regional localizers and then derive them from a well-trained classifier. This work has been accepted by ECCV2022. | Repository License |
| OSP-09 | CCLD-Net: Boosting medical image segmentation via conditional-synergistic convolution and lesion decoupling
| Date Created Owner Implement in | In this paper, we propose an input-specific network called conditional-synergistic convolution and lesion decoupling network (CCLDNet) to solve these issues. Our CCLDNet outperforms state-of-the-art approaches by a large margin on a variety of benchmarks. The related work has been published by Computerized Medical Imaging and Graphics. . | Repository License Related Paper |
| OSP-08 | NI‐P3D‐Unet OCTA
| Date Created Owner Implement in | In this paper, we designed a neighborhood information‐fused Pseudo‐3D U‐Net (NI‐P3D‐U) for OCTA reconstruction. The proposed NI-P3D-U was investigated on an in vivo animal dataset by a cross-validation strategy under both fully supervised learning and weakly supervised learning pipelines. The related work has been published by Medical Physics. | Data License Related Paper |
| OSP-07 | BLRM: Bayesian Statistics Guided Label Refurbishment Mechanism
| Date Created Owner Implement in | In this work, we proposed a novel Bayesian statistics guided label refurbishment mechanism (BLRM) for DNNs to prevent overftting noisy images. BLRM utilized maximum a posteriori probability (MAP) in the Bayesian statistics and the exponentially time-weighted technique to selectively correct the labels of noisy images. This work has been accepted by Medical Physics. | Repository License Related Paper |
| OSP-06 | DA-WSOL | Date Created Owner Implement in | DA-WSOL demonstrates the implementation of paper ``Weakly Supervised Object Localization as Domain Adaption". Our method views WSOL as adapting the image-level features and pixel-level features, which is respectively focused by training and testing process. Moreover, our work also provide an efficient pipeline to utilize DA methods for assisting different of WSOL methods by proposing a DAL loss and a target sample assigner. This work has been accepted by CVPR2022. | Repository License Related Paper |
| OSP-05 | TCFL_Unpaired_OCT_Denoising
| Date Created Owner Implement in | TCFL_Unpaired_OCT_Denoising demonstrates the implementation of the Triplet Cross-Fusion Learning for Unpaired Image Denoising in Optical Coherence Tomography. The proposed TCFL strategy is able to utilize concise structure to effectively reduce speckle noise, instead of designing complex network structures which is time-consuming and requires extensive empirical knowledge. This work has been accepted by IEEE Transactions on Medical Imaging. | Repository Data License Related Paper |
| OSP-04 | Spectral Nonlocal Block
| Date Created Owner Implement in | The Spectral View of Non-local (SNL) provide a novel perspective for the model design of non-local blocks. Our spectral view can help to theoretically analyze exsiting non-local blocks and design novel non-local block with the help of graph signal processing (e.g. the graph neural networks). This work has been published by the Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV2021) | Repository License Related Paper |
| OSP-03 | CNCL_Medical_Image_Denoising
| Date Created Owner Implement in | CNCL_Medical_Image_Denoising demonstrates the implementation of the Content-Noise Complementary Learning for Medical Imaging Denoising. This denoising pipeline processes both the content and the noise in the denoising task, and has shown generalization ability in many types of medical imaging. This work has been accepted by IEEE Transactions on Medical Imaging. | Repository License Related Paper |
| OSP-02 | Weakly Supervised DL-OCTA
| Date Created Owner Implement in | In this paper, we proposed a weakly supervised deep learning-based pipeline for OCTA reconstruction task, in the absence of high-quality training labels.The proposed pipeline was investigated on an in-vivo animal dataset and a human eye dataset by a crossvalidationstrategy. The related work has been published by IEEE Transactions on Medical Imaging. | Repository Data License Related Paper |
| OSP-01 | PMRMC
| Date Created Owner Contribute by Implement in | PMRMC is a Monte Carlo simulation of the transportation and annihilation process of positrons inside a uniform magnetic field. PMRMC was created by Xiangxi Meng in Beijing Cancer Hospital. Lujia Jin (PhD student in MILab) also made a significant contribution. The related work has been published by Medical Physics. | Repositroy License Related Paper |
Other open source projects
May include: open dataset, shared protocol, and etc.