Search for: All records

Wen, S., Wang, H., Metaxas, D. (2022). Social ODE: Multi-agent Trajectory Forecasting with Neural Ordinary Differential Equations. In: Avidan, S., Brostow, G., Cissé, M., Farinella, G.M., Hassner, T. (eds) Computer Vision – ECCV 2022. ECCV 2022. Lecture Notes in Computer Science, vol 13682. Springer, Cham. https://doi.org/10.1007/978-3-031-20047-2_13 Multi-agent trajectory forecasting has recently attracted a lot of attention due to its widespread applications including autonomous driving. Most previous methods use RNNs or Transformers to model agent dynamics in the temporal dimension and social pooling or GNNs to model interactions with other agents; these approaches usually fail to learn the underlying continuous temporal dynamics and agent interactions explicitly. To address these problems, we propose Social ODE which explicitly models temporal agent dynamics and agent interactions. Our approach leverages Neural ODEs to model continuous temporal dynamics, and incorporates distance, interaction intensity, and aggressiveness estimation into agent interaction modeling in latent space. We show in extensive experiments that our Social ODE approach compares favorably with state-of-the-art, and more importantly, can successfully avoid sudden obstacles and effectively control the motion of the agent, while previous methods often fail in such cases. 

Manual examination of chest x-rays is a time consuming process that involves significant effort by expert radiologists. Recent work attempts to alleviate this problem by developing learning-based automated chest x-ray analysis systems that map images to multi-label diagnoses using deep neural net- works. These methods are often treated as black boxes, or they output attention maps but don’t explain why the attended areas are important. Given data consisting of a frontal-view x-ray, a set of natural language findings, and one or more diagnostic impressions, we propose a deep neural network model that during training simultaneously 1) constructs a topic model which clusters key terms from the findings into meaningful groups, 2) predicts the presence of each topic for a given input image based on learned visual features, and 3) uses an image’s predicted topic encoding as features to predict one or more diagnoses. Since the net learns the topic model jointly with the classifier, it gives us a powerful tool for understanding which semantic concepts the net might be ex- ploiting when making diagnoses, and since we constrain the net to predict topics based on expert-annotated reports, the net automatically encodes some higher-level expert knowledge about how to make diagnoses. 

Temporal correlation in dynamic magnetic resonance imaging (MRI), such as cardiac MRI, is in- formative and important to understand motion mechanisms of body regions. Modeling such in- formation into the MRI reconstruction process produces temporally coherent image sequence and reduces imaging artifacts and blurring. However, existing deep learning based approaches neglect motion information during the reconstruction procedure, while traditional motion-guided methods are hindered by heuristic parameter tuning and long inference time. We propose a novel dynamic MRI reconstruction approach called MODRN that unitizes deep neural networks with motion in- formation to improve reconstruction quality. The central idea is to decompose the motion-guided optimization problem of dynamic MRI reconstruction into three components: dynamic reconstruc- tion, motion estimation and motion compensation. Extensive experiments have demonstrated the effectiveness of our proposed approach compared to other state-of-the-art approaches. 

Existing neural cell tracking methods generally use the morphology cell features for data association. However, these features are limited to the quality of cell segmentation and are prone to errors for mitosis determination. To over- come these issues, in this work we propose an online multi- object tracking method that leverages both cell appearance and motion features for data association. In particular, we propose a supervised blob-seed network (BSNet) to predict the cell appearance features and an unsupervised optical flow network (UnFlowNet) for capturing the cell motions. The data association is then solved using the Hungarian al- gorithm. Experimental evaluation shows that our approach achieves better performance than existing neural cell track- ing methods. 

Nuclei segmentation is a fundamental task in histopathological image analysis. Typically, such segmentation tasks require significant effort to manually generate pixel-wise annotations for fully supervised training. To alleviate the manual effort, in this paper we propose a novel approach using points only annotation. Two types of coarse labels with complementary information are derived from the points annotation, and are then utilized to train a deep neural network. The fully- connected conditional random field loss is utilized to further refine the model without introducing extra computational complexity during inference. Experimental results on two nuclei segmentation datasets reveal that the proposed method is able to achieve competitive performance compared to the fully supervised counterpart and the state-of-the-art methods while requiring significantly less annotation effort. Our code is publicly available. 

We consider an MRI reconstruction problem with input of k-space data at a very low undersampled rate. This can prac- tically benefit patient due to reduced time of MRI scan, but it is also challenging since quality of reconstruction may be compromised. Currently, deep learning based methods dom- inate MRI reconstruction over traditional approaches such as Compressed Sensing, but they rarely show satisfactory performance in the case of low undersampled k-space data. One explanation is that these methods treat channel-wise fea- tures equally, which results in degraded representation ability of the neural network. To solve this problem, we propose a new model called MRI Cascaded Channel-wise Attention Network (MICCAN), highlighted by three components: (i) a variant of U-net with Channel-wise Attention (UCA) mod- ule, (ii) a long skip connection and (iii) a combined loss. Our model is able to attend to salient information by filtering irrelevant features and also concentrate on high-frequency in- formation by enforcing low-frequency information bypassed to the final output. We conduct both quantitative evaluation and qualitative analysis of our method on a cardiac dataset. The experiment shows that our method achieves very promis- ing results in terms of three common metrics on the MRI reconstruction with low undersampled k-space data. Code is public available 

Neural cell instance segmentation serves as a valuable tool for the study of neural cell behaviors. In general, the instance segmentation methods compute the region of interest (ROI) through a detection module, where the segmentation is sub- sequently performed. To precisely segment the neural cells, especially their tiny and slender structures, existing work em- ploys a u-net structure to preserve the low-level details and encode the high-level semantics. However, such method is insufficient for differentiating the adjacent cells when large parts of them are included in the same cropped ROI. To solve this problem, we propose a context-refined neural cell instance segmentation model that learns to suppress the back- ground information. In particular, we employ a light-weight context refinement module to recalibrate the deep features and focus the model exclusively on the target cell within each cropped ROI. The proposed model is efficient and accurate, and experimental results demonstrate its superiority com- pared to the state-of-the-arts. 

Nuclei segmentation and classification are two important tasks in the histopathology image analysis, because the mor- phological features of nuclei and spatial distributions of dif- ferent types of nuclei are highly related to cancer diagnosis and prognosis. Existing methods handle the two problems independently, which are not able to obtain the features and spatial heterogeneity of different types of nuclei at the same time. In this paper, we propose a novel deep learning based method which solves both tasks in a unified framework. It can segment individual nuclei and classify them into tumor, lymphocyte and stroma nuclei. Perceptual loss is utilized to enhance the segmentation of details. We also take advantages of transfer learning to promote the training of deep neural net- works on a relatively small lung cancer dataset. Experimental results prove the effectiveness of the proposed method. The code is publicly available 

The ability for computational agents to reason about the high-level content of real world scene images is important for many applications. Existing attempts at complex scene understanding lack representational power, efficiency, and the ability to create robust meta- knowledge about scenes. We introduce scenarios as a new way of representing scenes. The scenario is an interpretable, low-dimensional, data-driven representation consisting of sets of frequently co-occurring objects that is useful for a wide range of scene under- standing tasks. Scenarios are learned from data using a novel matrix factorization method which is integrated into a new neural network architecture, the Scenari-oNet. Using ScenarioNet, we can recover semantic in- formation about real world scene images at three levels of granularity: 1) scene categories, 2) scenarios, and 3) objects. Training a single ScenarioNet model enables us to perform scene classification, scenario recognition, multi-object recognition, content-based scene image retrieval, and content-based image comparison. ScenarioNet is efficient because it requires significantly fewer parameters than other CNNs while achieving similar performance on benchmark tasks, and it is interpretable because it produces evidence in an understandable format for every decision it makes. We validate the utility of scenarios and ScenarioNet on a diverse set of scene understanding tasks on several benchmark datasets.