Kashu Yamazaki

Kashu Yamazaki, Khoa Vo, Sang Truong et al. · cmu

Video paragraph captioning aims to generate a multi-sentence description of an untrimmed video with several temporal event locations in coherent storytelling. Following the human perception process, where the scene is effectively understood by decomposing it into visual (e.g. human, animal) and non-visual components (e.g. action, relations) under the mutual influence of vision and language, we first propose a visual-linguistic (VL) feature. In the proposed VL feature, the scene is modeled by three modalities including (i) a global visual environment; (ii) local visual main agents; (iii) linguistic scene elements. We then introduce an autoregressive Transformer-in-Transformer (TinT) to simultaneously capture the semantic coherence of intra- and inter-event contents within a video. Finally, we present a new VL contrastive loss function to guarantee learnt embedding features are matched with the captions semantics. Comprehensive experiments and extensive ablation studies on ActivityNet Captions and YouCookII datasets show that the proposed Visual-Linguistic Transformer-in-Transform (VLTinT) outperforms prior state-of-the-art methods on accuracy and diversity. Source code is made publicly available at: https://github.com/UARK-AICV/VLTinT.

Khoa Vo, Sang Truong, Kashu Yamazaki et al. · cmu

Temporal action proposal generation (TAPG) is a challenging task, which requires localizing action intervals in an untrimmed video. Intuitively, we as humans, perceive an action through the interactions between actors, relevant objects, and the surrounding environment. Despite the significant progress of TAPG, a vast majority of existing methods ignore the aforementioned principle of the human perceiving process by applying a backbone network into a given video as a black-box. In this paper, we propose to model these interactions with a multi-modal representation network, namely, Actors-Objects-Environment Interaction Network (AOE-Net). Our AOE-Net consists of two modules, i.e., perception-based multi-modal representation (PMR) and boundary-matching module (BMM). Additionally, we introduce adaptive attention mechanism (AAM) in PMR to focus only on main actors (or relevant objects) and model the relationships among them. PMR module represents each video snippet by a visual-linguistic feature, in which main actors and surrounding environment are represented by visual information, whereas relevant objects are depicted by linguistic features through an image-text model. BMM module processes the sequence of visual-linguistic features as its input and generates action proposals. Comprehensive experiments and extensive ablation studies on ActivityNet-1.3 and THUMOS-14 datasets show that our proposed AOE-Net outperforms previous state-of-the-art methods with remarkable performance and generalization for both TAPG and temporal action detection. To prove the robustness and effectiveness of AOE-Net, we further conduct an ablation study on egocentric videos, i.e. EPIC-KITCHENS 100 dataset. Source code is available upon acceptance.

Hyekang Kevin Joo, Khoa Vo, Kashu Yamazaki et al. · cmu

Video anomaly detection (VAD) -- commonly formulated as a multiple-instance learning problem in a weakly-supervised manner due to its labor-intensive nature -- is a challenging problem in video surveillance where the frames of anomaly need to be localized in an untrimmed video. In this paper, we first propose to utilize the ViT-encoded visual features from CLIP, in contrast with the conventional C3D or I3D features in the domain, to efficiently extract discriminative representations in the novel technique. We then model temporal dependencies and nominate the snippets of interest by leveraging our proposed Temporal Self-Attention (TSA). The ablation study confirms the effectiveness of TSA and ViT feature. The extensive experiments show that our proposed CLIP-TSA outperforms the existing state-of-the-art (SOTA) methods by a large margin on three commonly-used benchmark datasets in the VAD problem (UCF-Crime, ShanghaiTech Campus, and XD-Violence). Our source code is available at https://github.com/joos2010kj/CLIP-TSA.

Minh Tran, Khoa Vo, Kashu Yamazaki et al. · cmu

Amodal Instance Segmentation (AIS) aims to segment the region of both visible and possible occluded parts of an object instance. While Mask R-CNN-based AIS approaches have shown promising results, they are unable to model high-level features coherence due to the limited receptive field. The most recent transformer-based models show impressive performance on vision tasks, even better than Convolution Neural Networks (CNN). In this work, we present AISFormer, an AIS framework, with a Transformer-based mask head. AISFormer explicitly models the complex coherence between occluder, visible, amodal, and invisible masks within an object's regions of interest by treating them as learnable queries. Specifically, AISFormer contains four modules: (i) feature encoding: extract ROI and learn both short-range and long-range visual features. (ii) mask transformer decoding: generate the occluder, visible, and amodal mask query embeddings by a transformer decoder (iii) invisible mask embedding: model the coherence between the amodal and visible masks, and (iv) mask predicting: estimate output masks including occluder, visible, amodal and invisible. We conduct extensive experiments and ablation studies on three challenging benchmarks i.e. KINS, D2SA, and COCOA-cls to evaluate the effectiveness of AISFormer. The code is available at: https://github.com/UARK-AICV/AISFormer

Khoa Vo, Kashu Yamazaki, Phong X. Nguyen et al. · cmu

Video understanding is a growing field and a subject of intense research, which includes many interesting tasks to understanding both spatial and temporal information, e.g., action detection, action recognition, video captioning, video retrieval. One of the most challenging problems in video understanding is dealing with feature extraction, i.e. extract contextual visual representation from given untrimmed video due to the long and complicated temporal structure of unconstrained videos. Different from existing approaches, which apply a pre-trained backbone network as a black-box to extract visual representation, our approach aims to extract the most contextual information with an explainable mechanism. As we observed, humans typically perceive a video through the interactions between three main factors, i.e., the actors, the relevant objects, and the surrounding environment. Therefore, it is very crucial to design a contextual explainable video representation extraction that can capture each of such factors and model the relationships between them. In this paper, we discuss approaches, that incorporate the human perception process into modeling actors, objects, and the environment. We choose video paragraph captioning and temporal action detection to illustrate the effectiveness of human perception based-contextual representation in video understanding. Source code is publicly available at https://github.com/UARK-AICV/Video_Representation.

Khoa Vo, Kashu Yamazaki, Sang Truong et al. · cmu

Temporal action proposal generation (TAPG) aims to estimate temporal intervals of actions in untrimmed videos, which is a challenging yet plays an important role in many tasks of video analysis and understanding. Despite the great achievement in TAPG, most existing works ignore the human perception of interaction between agents and the surrounding environment by applying a deep learning model as a black-box to the untrimmed videos to extract video visual representation. Therefore, it is beneficial and potentially improve the performance of TAPG if we can capture these interactions between agents and the environment. In this paper, we propose a novel framework named Agent-Aware Boundary Network (ABN), which consists of two sub-networks (i) an Agent-Aware Representation Network to obtain both agent-agent and agents-environment relationships in the video representation, and (ii) a Boundary Generation Network to estimate the confidence score of temporal intervals. In the Agent-Aware Representation Network, the interactions between agents are expressed through local pathway, which operates at a local level to focus on the motions of agents whereas the overall perception of the surroundings are expressed through global pathway, which operates at a global level to perceive the effects of agents-environment. Comprehensive evaluations on 20-action THUMOS-14 and 200-action ActivityNet-1.3 datasets with different backbone networks (i.e C3D, SlowFast and Two-Stream) show that our proposed ABN robustly outperforms state-of-the-art methods regardless of the employed backbone network on TAPG. We further examine the proposal quality by leveraging proposals generated by our method onto temporal action detection (TAD) frameworks and evaluate their detection performances. The source code can be found in this URL https://github.com/vhvkhoa/TAPG-AgentEnvNetwork.git.

Kashu Yamazaki, Taisei Hanyu, Minh Tran et al. · cmu

Aerial Image Segmentation is a top-down perspective semantic segmentation and has several challenging characteristics such as strong imbalance in the foreground-background distribution, complex background, intra-class heterogeneity, inter-class homogeneity, and tiny objects. To handle these problems, we inherit the advantages of Transformers and propose AerialFormer, which unifies Transformers at the contracting path with lightweight Multi-Dilated Convolutional Neural Networks (MD-CNNs) at the expanding path. Our AerialFormer is designed as a hierarchical structure, in which Transformer encoder outputs multi-scale features and MD-CNNs decoder aggregates information from the multi-scales. Thus, it takes both local and global contexts into consideration to render powerful representations and high-resolution segmentation. We have benchmarked AerialFormer on three common datasets including iSAID, LoveDA, and Potsdam. Comprehensive experiments and extensive ablation studies show that our proposed AerialFormer outperforms previous state-of-the-art methods with remarkable performance. Our source code will be publicly available upon acceptance.

Kashu Yamazaki, Sang Truong, Khoa Vo et al. · cmu

In this paper, we leverage the human perceiving process, that involves vision and language interaction, to generate a coherent paragraph description of untrimmed videos. We propose vision-language (VL) features consisting of two modalities, i.e., (i) vision modality to capture global visual content of the entire scene and (ii) language modality to extract scene elements description of both human and non-human objects (e.g. animals, vehicles, etc), visual and non-visual elements (e.g. relations, activities, etc). Furthermore, we propose to train our proposed VLCap under a contrastive learning VL loss. The experiments and ablation studies on ActivityNet Captions and YouCookII datasets show that our VLCap outperforms existing SOTA methods on both accuracy and diversity metrics.

Kashu Yamazaki, Taisei Hanyu, Khoa Vo et al. · cmu

Precise 3D environmental mapping is pivotal in robotics. Existing methods often rely on predefined concepts during training or are time-intensive when generating semantic maps. This paper presents Open-Fusion, a groundbreaking approach for real-time open-vocabulary 3D mapping and queryable scene representation using RGB-D data. Open-Fusion harnesses the power of a pre-trained vision-language foundation model (VLFM) for open-set semantic comprehension and employs the Truncated Signed Distance Function (TSDF) for swift 3D scene reconstruction. By leveraging the VLFM, we extract region-based embeddings and their associated confidence maps. These are then integrated with 3D knowledge from TSDF using an enhanced Hungarian-based feature-matching mechanism. Notably, Open-Fusion delivers outstanding annotation-free 3D segmentation for open-vocabulary without necessitating additional 3D training. Benchmark tests on the ScanNet dataset against leading zero-shot methods highlight Open-Fusion's superiority. Furthermore, it seamlessly combines the strengths of region-based VLFM and TSDF, facilitating real-time 3D scene comprehension that includes object concepts and open-world semantics. We encourage the readers to view the demos on our project page: https://uark-aicv.github.io/OpenFusion

Viet-Khoa Vo-Ho, Kashu Yamazaki, Hieu Hoang et al. · cmu, microsoft-research

Deep learning methods have been successful in solving tasks in machine learning and have made breakthroughs in many sectors owing to their ability to automatically extract features from unstructured data. However, their performance relies on manual trial-and-error processes for selecting an appropriate network architecture, hyperparameters for training, and pre-/post-procedures. Even though it has been shown that network architecture plays a critical role in learning feature representation feature from data and the final performance, searching for the best network architecture is computationally intensive and heavily relies on researchers' experience. Automated machine learning (AutoML) and its advanced techniques i.e. Neural Architecture Search (NAS) have been promoted to address those limitations. Not only in general computer vision tasks, but NAS has also motivated various applications in multiple areas including medical imaging. In medical imaging, NAS has significant progress in improving the accuracy of image classification, segmentation, reconstruction, and more. However, NAS requires the availability of large annotated data, considerable computation resources, and pre-defined tasks. To address such limitations, meta-learning has been adopted in the scenarios of few-shot learning and multiple tasks. In this book chapter, we first present a brief review of NAS by discussing well-known approaches in search space, search strategy, and evaluation strategy. We then introduce various NAS approaches in medical imaging with different applications such as classification, segmentation, detection, reconstruction, etc. Meta-learning in NAS for few-shot learning and multiple tasks is then explained. Finally, we describe several open problems in NAS.

Ziyan Xiong, Lixing Fang, Junyun Huang et al. · cmu

Building a low-latency humanoid teleoperation system is essential for collecting diverse reactive and dynamic demonstrations. However, existing approaches rely on heavily pre-processed human-to-humanoid motion retargeting and position-only PD control, resulting in substantial latency that severely limits responsiveness and prevents tasks requiring rapid feedback and fast reactions. To address this problem, we propose ExtremControl, a low latency whole-body control framework that: (1) operates directly on SE(3) poses of selected rigid links, primarily humanoid extremities, to avoid full-body retargeting; (2) utilizes a Cartesian-space mapping to directly convert human motion to humanoid link targets; and (3) incorporates velocity feedforward control at low level to support highly responsive behavior under rapidly changing control interfaces. We further provide a unified theoretical formulation of ExtremControl and systematically validate its effectiveness through experiments in both simulation and real-world environments. Building on ExtremControl, we implement a low-latency humanoid teleoperation system that supports both optical motion capture and VR-based motion tracking, achieving end-to-end latency as low as 50ms and enabling highly responsive behaviors such as ping-pong ball balancing, juggling, and real-time return, thereby substantially surpassing the 200ms latency limit observed in prior work.

Yuto Shibata, Kashu Yamazaki, Lalit Jayanti et al. · cmu

Humanoid robotics has strong potential to transform daily service and caregiving applications. Although recent advances in general motion tracking within physics engines (GMT) have enabled virtual characters and humanoid robots to reproduce a broad range of human motions, these behaviors are primarily limited to contact-less social interactions or isolated movements. Assistive scenarios, by contrast, require continuous awareness of a human partner and rapid adaptation to their evolving posture and dynamics. In this paper, we formulate the imitation of closely interacting, force-exchanging human-human motion sequences as a multi-agent reinforcement learning problem. We jointly train partner-aware policies for both the supporter (assistant) agent and the recipient agent in a physics simulator to track assistive motion references. To make this problem tractable, we introduce a partner policies initialization scheme that transfers priors from single-human motion-tracking controllers, greatly improving exploration. We further propose dynamic reference retargeting and contact-promoting reward, which adapt the assistant's reference motion to the recipient's real-time pose and encourage physically meaningful support. We show that AssistMimic is the first method capable of successfully tracking assistive interaction motions on established benchmarks, demonstrating the benefits of a multi-agent RL formulation for physically grounded and socially aware humanoid control.

Nhat Chung, Taisei Hanyu, Toan Nguyen et al.

As embodied agents operate in increasingly complex environments, the ability to perceive, track, and reason about individual object instances over time becomes essential, especially in tasks requiring sequenced interactions with visually similar objects. In these non-Markovian settings, key decision cues are often hidden in object-specific histories rather than the current scene. Without persistent memory of prior interactions (what has been interacted with, where it has been, or how it has changed) visuomotor policies may fail, repeat past actions, or overlook completed ones. To surface this challenge, we introduce LIBERO-Mem, a non-Markovian task suite for stress-testing robotic manipulation under object-level partial observability. It combines short- and long-horizon object tracking with temporally sequenced subgoals, requiring reasoning beyond the current frame. However, vision-language-action (VLA) models often struggle in such settings, with token scaling quickly becoming intractable even for tasks spanning just a few hundred frames. We propose Embodied-SlotSSM, a slot-centric VLA framework built for temporal scalability. It maintains spatio-temporally consistent slot identities and leverages them through two mechanisms: (1) slot-state-space modeling for reconstructing short-term history, and (2) a relational encoder to align the input tokens with action decoding. Together, these components enable temporally grounded, context-aware action prediction. Experiments show Embodied-SlotSSM's baseline performance on LIBERO-Mem and general tasks, offering a scalable solution for non-Markovian reasoning in object-centric robotic policies.

Taisei Hanyu, Nhat Chung, Huy Le et al.

Inspired by how humans reason over discrete objects and their relationships, we explore whether compact object-centric and object-relation representations can form a foundation for multitask robotic manipulation. Most existing robotic multitask models rely on dense embeddings that entangle both object and background cues, raising concerns about both efficiency and interpretability. In contrast, we study object-relation-centric representations as a pathway to more structured, efficient, and explainable visuomotor control. Our contributions are two-fold. First, we introduce LIBERO+, a fine-grained benchmark dataset designed to enable and evaluate object-relation reasoning in robotic manipulation. Unlike prior datasets, LIBERO+ provides object-centric annotations that enrich demonstrations with box- and mask-level labels as well as instance-level temporal tracking, supporting compact and interpretable visuomotor representations. Second, we propose SlotVLA, a slot-attention-based framework that captures both objects and their relations for action decoding. It uses a slot-based visual tokenizer to maintain consistent temporal object representations, a relation-centric decoder to produce task-relevant embeddings, and an LLM-driven module that translates these embeddings into executable actions. Experiments on LIBERO+ demonstrate that object-centric slot and object-relation slot representations drastically reduce the number of required visual tokens, while providing competitive generalization. Together, LIBERO+ and SlotVLA provide a compact, interpretable, and effective foundation for advancing object-relation-centric robotic manipulation.

Yash Jangir, Yidi Zhang, Kashu Yamazaki et al.

The pursuit of robot generalists - instructable agents capable of performing diverse tasks across diverse environments - demands rigorous and scalable evaluation. Yet real-world testing of robot policies remains fundamentally constrained: it is labor-intensive, slow, unsafe at scale, and difficult to reproduce. Existing simulation benchmarks are similarly limited, as they train and test policies within the same synthetic domains and cannot assess models trained from real-world demonstrations or alternative simulation environments. As policies expand in scope and complexity, these barriers only intensify, since defining "success" in robotics often hinges on nuanced human judgments of execution quality. In this paper, we introduce a new benchmarking framework that overcomes these challenges by shifting VLA evaluation into large-scale simulated environments augmented with online human feedback. Leveraging advances in vision-language models, 2D-to-3D generative modeling, and differentiable rendering, our approach automatically converts video demonstrations from widely used robot datasets into simulated counterparts. Within these digital twins, we assess VLA policies using both automated VLM-guided scoring and scalable human preference judgments collected from crowdworkers, transforming human involvement from tedious scene setup, resetting, and safety supervision into lightweight preference comparisons. To measure robustness, we systematically perturb simulated environments along multiple axes, such as textures and object placements, stress-testing policy generalization under controlled variation. The result is a continuously evolving, reproducible, and scalable benchmark for real-world trained robot manipulation policies, addressing a critical missing capability in today's robotics landscape.

Khoa Vo, Thinh Phan, Kashu Yamazaki et al.

Current video-language models (VLMs) rely extensively on instance-level alignment between video and language modalities, which presents two major limitations: (1) visual reasoning disobeys the natural perception that humans do in first-person perspective, leading to a lack of reasoning interpretation; and (2) learning is limited in capturing inherent fine-grained relationships between two modalities. In this paper, we take an inspiration from human perception and explore a compositional approach for egocentric video representation. We introduce HENASY (Hierarchical ENtities ASsemblY), which includes a spatiotemporal token grouping mechanism to explicitly assemble dynamically evolving scene entities through time and model their relationship for video representation. By leveraging compositional structure understanding, HENASY possesses strong interpretability via visual grounding with free-form text queries. We further explore a suite of multi-grained contrastive losses to facilitate entity-centric understandings. This comprises three alignment types: video-narration, noun-entity, verb-entities alignments. Our method demonstrates strong interpretability in both quantitative and qualitative experiments; while maintaining competitive performances on five downstream tasks via zero-shot transfer or as video/text representation, including video/text retrieval, action recognition, multi-choice query, natural language query, and moments query.

Khoa Vo, Hyekang Joo, Kashu Yamazaki et al.

Humans typically perceive the establishment of an action in a video through the interaction between an actor and the surrounding environment. An action only starts when the main actor in the video begins to interact with the environment, while it ends when the main actor stops the interaction. Despite the great progress in temporal action proposal generation, most existing works ignore the aforementioned fact and leave their model learning to propose actions as a black-box. In this paper, we make an attempt to simulate that ability of a human by proposing Actor Environment Interaction (AEI) network to improve the video representation for temporal action proposals generation. AEI contains two modules, i.e., perception-based visual representation (PVR) and boundary-matching module (BMM). PVR represents each video snippet by taking human-human relations and humans-environment relations into consideration using the proposed adaptive attention mechanism. Then, the video representation is taken by BMM to generate action proposals. AEI is comprehensively evaluated in ActivityNet-1.3 and THUMOS-14 datasets, on temporal action proposal and detection tasks, with two boundary-matching architectures (i.e., CNN-based and GCN-based) and two classifiers (i.e., Unet and P-GCN). Our AEI robustly outperforms the state-of-the-art methods with remarkable performance and generalization for both temporal action proposal generation and temporal action detection.

Ngan Le, Vidhiwar Singh Rathour, Kashu Yamazaki et al.

Deep reinforcement learning augments the reinforcement learning framework and utilizes the powerful representation of deep neural networks. Recent works have demonstrated the remarkable successes of deep reinforcement learning in various domains including finance, medicine, healthcare, video games, robotics, and computer vision. In this work, we provide a detailed review of recent and state-of-the-art research advances of deep reinforcement learning in computer vision. We start with comprehending the theories of deep learning, reinforcement learning, and deep reinforcement learning. We then propose a categorization of deep reinforcement learning methodologies and discuss their advantages and limitations. In particular, we divide deep reinforcement learning into seven main categories according to their applications in computer vision, i.e. (i)landmark localization (ii) object detection; (iii) object tracking; (iv) registration on both 2D image and 3D image volumetric data (v) image segmentation; (vi) videos analysis; and (vii) other applications. Each of these categories is further analyzed with reinforcement learning techniques, network design, and performance. Moreover, we provide a comprehensive analysis of the existing publicly available datasets and examine source code availability. Finally, we present some open issues and discuss future research directions on deep reinforcement learning in computer vision

Viet-Khoa Vo-Ho, Ngan Le, Kashu Yamazaki et al.

Temporal action proposal generation is an essential and challenging task that aims at localizing temporal intervals containing human actions in untrimmed videos. Most of existing approaches are unable to follow the human cognitive process of understanding the video context due to lack of attention mechanism to express the concept of an action or an agent who performs the action or the interaction between the agent and the environment. Based on the action definition that a human, known as an agent, interacts with the environment and performs an action that affects the environment, we propose a contextual Agent-Environment Network. Our proposed contextual AEN involves (i) agent pathway, operating at a local level to tell about which humans/agents are acting and (ii) environment pathway operating at a global level to tell about how the agents interact with the environment. Comprehensive evaluations on 20-action THUMOS-14 and 200-action ActivityNet-1.3 datasets with different backbone networks, i.e C3D and SlowFast, show that our method robustly exhibits outperformance against state-of-the-art methods regardless of the employed backbone network.

Kashu Yamazaki, Vidhiwar Singh Rathour, T. Hoang Ngan Le

Deep Convolutional Neural Networks (CNNs) i.e. Residual Networks (ResNets) have been used successfully for many computer vision tasks, but are difficult to scale to 3D volumetric medical data. Memory is increasingly often the bottleneck when training 3D Convolutional Neural Networks (CNNs). Recently, invertible neural networks have been applied to significantly reduce activation memory footprint when training neural networks with backpropagation thanks to the invertible functions that allow retrieving its input from its output without storing intermediate activations in memory to perform the backpropagation. Among many successful network architectures, 3D Unet has been established as a standard architecture for volumetric medical segmentation. Thus, we choose 3D Unet as a baseline for a non-invertible network and we then extend it with the invertible residual network. In this paper, we proposed two versions of the invertible Residual Network, namely Partially Invertible Residual Network (Partially-InvRes) and Fully Invertible Residual Network (Fully-InvRes). In Partially-InvRes, the invertible residual layer is defined by a technique called additive coupling whereas in Fully-InvRes, both invertible upsampling and downsampling operations are learned based on squeezing (known as pixel shuffle). Furthermore, to avoid the overfitting problem because of less training data, a variational auto-encoder (VAE) branch is added to reconstruct the input volumetric data itself. Our results indicate that by using partially/fully invertible networks as the central workhorse in volumetric segmentation, we not only reduce memory overhead but also achieve compatible segmentation performance compared against the non-invertible 3D Unet. We have demonstrated the proposed networks on various volumetric datasets such as iSeg 2019 and BraTS 2020.

Ngan Le, Trung Le, Kashu Yamazaki et al.

Medical image segmentation has played an important role in medical analysis and widely developed for many clinical applications. Deep learning-based approaches have achieved high performance in semantic segmentation but they are limited to pixel-wise setting and imbalanced classes data problem. In this paper, we tackle those limitations by developing a new deep learning-based model which takes into account both higher feature level i.e. region inside contour, intermediate feature level i.e. offset curves around the contour and lower feature level i.e. contour. Our proposed Offset Curves (OsC) loss consists of three main fitting terms. The first fitting term focuses on pixel-wise level segmentation whereas the second fitting term acts as attention model which pays attention to the area around the boundaries (offset curves). The third terms plays a role as regularization term which takes the length of boundaries into account. We evaluate our proposed OsC loss on both 2D network and 3D network. Two common medical datasets, i.e. retina DRIVE and brain tumor BRATS 2018 datasets are used to benchmark our proposed loss performance. The experiments have shown that our proposed OsC loss function outperforms other mainstream loss functions such as Cross-Entropy, Dice, Focal on the most common segmentation networks Unet, FCN.

Ngan Le, Kashu Yamazaki, Dat Truong et al.

In recent years, deep neural networks have achieved state-of-the-art performance in a variety of recognition and segmentation tasks in medical imaging including brain tumor segmentation. We investigate that segmenting a brain tumor is facing to the imbalanced data problem where the number of pixels belonging to the background class (non tumor pixel) is much larger than the number of pixels belonging to the foreground class (tumor pixel). To address this problem, we propose a multi-task network which is formed as a cascaded structure. Our model consists of two targets, i.e., (i) effectively differentiate the brain tumor regions and (ii) estimate the brain tumor mask. The first objective is performed by our proposed contextual brain tumor detection network, which plays a role of an attention gate and focuses on the region around brain tumor only while ignoring the far neighbor background which is less correlated to the tumor. The second objective is built upon a 3D atrous residual network and under an encode-decode network in order to effectively segment both large and small objects (brain tumor). Our 3D atrous residual network is designed with a skip connection to enables the gradient from the deep layers to be directly propagated to shallow layers, thus, features of different depths are preserved and used for refining each other. In order to incorporate larger contextual information from volume MRI data, our network utilizes the 3D atrous convolution with various kernel sizes, which enlarges the receptive field of filters. Our proposed network has been evaluated on various datasets including BRATS2015, BRATS2017 and BRATS2018 datasets with both validation set and testing set. Our performance has been benchmarked by both region-based metrics and surface-based metrics. We also have conducted comparisons against state-of-the-art approaches.