A comprehensive understanding of surgical scenes allows for monitoring of the surgical process, reducing the occurrence of accidents and enhancing efficiency for medical professionals. Semantic modeling within operating rooms, as a scene graph generation (SGG) task, is challenging since it involves consecutive recognition of subtle surgical actions over prolonged periods. To address this challenge, we propose a Tri-modal (i.e., images, point clouds, and language) confluence with Temporal dynamics framework, termed TriTemp-OR. Diverging from previous approaches that integrated temporal information via memory graphs, our method embraces two advantages: 1) we directly exploit bi-modal temporal information from the video streaming for hierarchical feature interaction, and 2) the prior knowledge from Large Language Models (LLMs) is embedded to alleviate the class-imbalance problem in the operating theatre. Specifically, our model performs temporal interactions across 2D frames and 3D point clouds, including a scale-adaptive multi-view temporal interaction (ViewTemp) and a geometric-temporal point aggregation (PointTemp). Furthermore, we transfer knowledge from the biomedical LLM, LLaVA-Med, to deepen the comprehension of intraoperative relations. The proposed TriTemp-OR enables the aggregation of tri-modal features through relation-aware unification to predict relations so as to generate scene graphs. Experimental results on the 4D-OR benchmark demonstrate the superior performance of our model for long-term OR streaming.
Scene graph generation (SGG) of surgical procedures is crucial in enhancing holistically cognitive intelligence in the operating room (OR). However, previous works have primarily relied on the multi-stage learning that generates semantic scene graphs dependent on intermediate processes with pose estimation and object detection, which may compromise model efficiency and efficacy, also impose extra annotation burden. In this study, we introduce a novel single-stage bimodal transformer framework for SGG in the OR, termed S^2Former-OR, aimed to complementally leverage multi-view 2D scenes and 3D point clouds for SGG in an end-to-end manner. Concretely, our model embraces a View-Sync Transfusion scheme to encourage multi-view visual information interaction. Concurrently, a Geometry-Visual Cohesion operation is designed to integrate the synergic 2D semantic features into 3D point cloud features. Moreover, based on the augmented feature, we propose a novel relation-sensitive transformer decoder that embeds dynamic entity-pair queries and relational trait priors, which enables the direct prediction of entity-pair relations for graph generation without intermediate steps. Extensive experiments have validated the superior SGG performance and lower computational cost of S^2Former-OR on 4D-OR benchmark, compared with current OR-SGG methods, e.g., 3% Precision increase and 24.2M reduction in model parameters. We further compared our method with generic single-stage SGG methods with broader metrics for a comprehensive evaluation, with consistently better performance achieved. The code will be made available.
The estimation of implicit cross-frame correspondences and the high computational cost have long been major challenges in video semantic segmentation (VSS) for driving scenes. Prior works utilize keyframes, feature propagation, or cross-frame attention to address these issues. By contrast, we are the first to harness vanishing point (VP) priors for more effective segmentation. Intuitively, objects near VPs (i.e., away from the vehicle) are less discernible. Moreover, they tend to move radially away from the VP over time in the usual case of a forward-facing camera, a straight road, and linear forward motion of the vehicle. Our novel, efficient network for VSS, named VPSeg, incorporates two modules that utilize exactly this pair of static and dynamic VP priors: sparse-to-dense feature mining (DenseVP) and VP-guided motion fusion (MotionVP). MotionVP employs VP-guided motion estimation to establish explicit correspondences across frames and help attend to the most relevant features from neighboring frames, while DenseVP enhances weak dynamic features in distant regions around VPs. These modules operate within a context-detail framework, which separates contextual features from high-resolution local features at different input resolutions to reduce computational costs. Contextual and local features are integrated through contextualized motion attention (CMA) for the final prediction. Extensive experiments on two popular driving segmentation benchmarks, Cityscapes and ACDC, demonstrate that VPSeg outperforms previous SOTA methods, with only modest computational overhead.
Data efficiency, or the ability to generalize from a few labeled data, remains a major challenge in deep learning. Semi-supervised learning has thrived in traditional recognition tasks alleviating the need for large amounts of labeled data, yet it remains understudied in image-to-image translation (I2I) tasks. In this work, we introduce the first semi-supervised (semi-paired) framework for label-to-image translation, a challenging subtask of I2I which generates photorealistic images from semantic label maps. In the semi-paired setting, the model has access to a small set of paired data and a larger set of unpaired images and labels. Instead of using geometrical transformations as a pretext task like previous works, we leverage an input reconstruction task by exploiting the conditional discriminator on the paired data as a reverse generator. We propose a training algorithm for this shared network, and we present a rare classes sampling algorithm to focus on under-represented classes. Experiments on 3 standard benchmarks show that the proposed model outperforms state-of-the-art unsupervised and semi-supervised approaches, as well as some fully supervised approaches while using a much smaller number of paired samples.
The need for large amounts of training and validation data is a huge concern in scaling AI algorithms for autonomous driving. Semantic Image Synthesis (SIS), or label-to-image translation, promises to address this issue by translating semantic layouts to images, providing a controllable generation of photorealistic data. However, they require a large amount of paired data, incurring extra costs. In this work, we present a new task: given a dataset with synthetic images and labels and a dataset with unlabeled real images, our goal is to learn a model that can generate images with the content of the input mask and the appearance of real images. This new task reframes the well-known unsupervised SIS task in a more practical setting, where we leverage cheaply available synthetic data from a driving simulator to learn how to generate photorealistic images of urban scenes. This stands in contrast to previous works, which assume that labels and images come from the same domain but are unpaired during training. We find that previous unsupervised works underperform on this task, as they do not handle distribution shifts between two different domains. To bypass these problems, we propose a novel framework with two main contributions. First, we leverage the synthetic image as a guide to the content of the generated image by penalizing the difference between their high-level features on a patch level. Second, in contrast to previous works which employ one discriminator that overfits the target domain semantic distribution, we employ a discriminator for the whole image and multiscale discriminators on the image patches. Extensive comparisons on the benchmarks GTA-V $\rightarrow$ Cityscapes and GTA-V $\rightarrow$ Mapillary show the superior performance of the proposed model against state-of-the-art on this task.