In the evolving field of Explainable AI (XAI), interpreting the decisions of deep neural networks (DNNs) in computer vision tasks is an important process. While pixel-based XAI methods focus on identifying significant pixels, existing concept-based XAI methods use pre-defined or human-annotated concepts. The recently proposed Segment Anything Model (SAM) achieved a significant step forward to prepare automatic concept sets via comprehensive instance segmentation. Building upon this, the Explain Any Concept (EAC) model emerged as a flexible method for explaining DNN decisions. EAC model is based on using a surrogate model which has one trainable linear layer to simulate the target model. In this paper, by introducing an additional nonlinear layer to the original surrogate model, we show that we can improve the performance of the EAC model. We compare our proposed approach to the original EAC model and report improvements obtained on both ImageNet and MS COCO datasets.
Instance segmentation is crucial for autonomous driving but is hindered by the lack of annotated real-world data due to expensive labeling costs. Unsupervised Domain Adaptation (UDA) offers a solution by transferring knowledge from labeled synthetic data to unlabeled real-world data. While UDA methods for synthetic to real-world domains (synth-to-real) show remarkable performance in tasks such as semantic segmentation and object detection, very few have been proposed for instance segmentation in vision-based autonomous driving. Moreover, existing methods rely on suboptimal baselines, which severely limits performance. We introduce \textbf{UDA4Inst}, a powerful framework for synth-to-real UDA in instance segmentation. Our framework enhances instance segmentation through \textit{Semantic Category Training} and \textit{Bidirectional Mixing Training}. With the Semantic Category Training method, semantically related classes are grouped and trained separately, enabling the generation of higher-quality pseudo-labels and improved segmentation performance. We further propose a bidirectional cross-domain data mixing strategy that combines instance-wise and patch-wise mixing techniques to effectively utilize data from both source and target domains, producing realistic composite images that improve the model's generalization performance. Extensive experiments demonstrate the effectiveness of our methods. Our approach establishes a new state-of-the-art on the SYNTHIA->Cityscapes benchmark with mAP 31.3. Notably, we are the first to report results on multiple novel synth-to-real instance segmentation datasets, using UrbanSyn and Synscapes as source domains while Cityscapes and KITTI360 serve as target domains. Our code will be released soon.
Instance segmentation is data-hungry, and as model capacity increases, data scale becomes crucial for improving the accuracy. Most instance segmentation datasets today require costly manual annotation, limiting their data scale. Models trained on such data are prone to overfitting on the training set, especially for those rare categories. While recent works have delved into exploiting generative models to create synthetic datasets for data augmentation, these approaches do not efficiently harness the full potential of generative models. To address these issues, we introduce a more efficient strategy to construct generative datasets for data augmentation, termed DiverGen. Firstly, we provide an explanation of the role of generative data from the perspective of distribution discrepancy. We investigate the impact of different data on the distribution learned by the model. We argue that generative data can expand the data distribution that the model can learn, thus mitigating overfitting. Additionally, we find that the diversity of generative data is crucial for improving model performance and enhance it through various strategies, including category diversity, prompt diversity, and generative model diversity. With these strategies, we can scale the data to millions while maintaining the trend of model performance improvement. On the LVIS dataset, DiverGen significantly outperforms the strong model X-Paste, achieving +1.1 box AP and +1.1 mask AP across all categories, and +1.9 box AP and +2.5 mask AP for rare categories.
Numerous applications have resulted from the automation of agricultural disease segmentation using deep learning techniques. However, when applied to new conditions, these applications frequently face the difficulty of overfitting, resulting in lower segmentation performance. In the context of potato farming, where diseases have a large influence on yields, it is critical for the agricultural economy to quickly and properly identify these diseases. Traditional data augmentation approaches, such as rotation, flip, and translation, have limitations and frequently fail to provide strong generalization results. To address these issues, our research employs a novel approach termed as PotatoGANs. In this novel data augmentation approach, two types of Generative Adversarial Networks (GANs) are utilized to generate synthetic potato disease images from healthy potato images. This approach not only expands the dataset but also adds variety, which helps to enhance model generalization. Using the Inception score as a measure, our experiments show the better quality and realisticness of the images created by PotatoGANs, emphasizing their capacity to resemble real disease images closely. The CycleGAN model outperforms the Pix2Pix GAN model in terms of image quality, as evidenced by its higher IS scores CycleGAN achieves higher Inception scores (IS) of 1.2001 and 1.0900 for black scurf and common scab, respectively. This synthetic data can significantly improve the training of large neural networks. It also reduces data collection costs while enhancing data diversity and generalization capabilities. Our work improves interpretability by combining three gradient-based Explainable AI algorithms (GradCAM, GradCAM++, and ScoreCAM) with three distinct CNN architectures (DenseNet169, Resnet152 V2, InceptionResNet V2) for potato disease classification.
In this paper, we show that transferring knowledge from other domains of video understanding combined with large-scale learning can improve robustness of Video Object Segmentation (VOS) under complex circumstances. Namely, we focus on integrating scene global motion knowledge to improve large-scale semi-supervised Video Object Segmentation. Prior works on VOS mostly rely on direct comparison of semantic and contextual features to perform dense matching between current and past frames, passing over actual motion structure. On the other hand, Optical Flow Estimation task aims to approximate the scene motion field, exposing global motion patterns which are typically undiscoverable during all pairs similarity search. We present WarpFormer, an architecture for semi-supervised Video Object Segmentation that exploits existing knowledge in motion understanding to conduct smoother propagation and more accurate matching. Our framework employs a generic pretrained Optical Flow Estimation network whose prediction is used to warp both past frames and instance segmentation masks to the current frame domain. Consequently, warped segmentation masks are refined and fused together aiming to inpaint occluded regions and eliminate artifacts caused by flow field imperfects. Additionally, we employ novel large-scale MOSE 2023 dataset to train model on various complex scenarios. Our method demonstrates strong performance on DAVIS 2016/2017 validation (93.0% and 85.9%), DAVIS 2017 test-dev (80.6%) and YouTube-VOS 2019 validation (83.8%) that is competitive with alternative state-of-the-art methods while using much simpler memory mechanism and instance understanding logic.
Pathology is the study of microscopic inspection of tissue, and a pathology diagnosis is often the medical gold standard to diagnose disease. Pathology images provide a unique challenge for computer-vision-based analysis: a single pathology Whole Slide Image (WSI) is gigapixel-sized and often contains hundreds of thousands to millions of objects of interest across multiple resolutions. In this work, we propose PathoLogy Universal TransfOrmer (PLUTO): a light-weight pathology FM that is pre-trained on a diverse dataset of 195 million image tiles collected from multiple sites and extracts meaningful representations across multiple WSI scales that enable a large variety of downstream pathology tasks. In particular, we design task-specific adaptation heads that utilize PLUTO's output embeddings for tasks which span pathology scales ranging from subcellular to slide-scale, including instance segmentation, tile classification, and slide-level prediction. We compare PLUTO's performance to other state-of-the-art methods on a diverse set of external and internal benchmarks covering multiple biologically relevant tasks, tissue types, resolutions, stains, and scanners. We find that PLUTO matches or outperforms existing task-specific baselines and pathology-specific foundation models, some of which use orders-of-magnitude larger datasets and model sizes when compared to PLUTO. Our findings present a path towards a universal embedding to power pathology image analysis, and motivate further exploration around pathology foundation models in terms of data diversity, architectural improvements, sample efficiency, and practical deployability in real-world applications.
Vision graph neural networks (ViG) offer a new avenue for exploration in computer vision. A major bottleneck in ViGs is the inefficient k-nearest neighbor (KNN) operation used for graph construction. To solve this issue, we propose a new method for designing ViGs, Dynamic Axial Graph Construction (DAGC), which is more efficient than KNN as it limits the number of considered graph connections made within an image. Additionally, we propose a novel CNN-GNN architecture, GreedyViG, which uses DAGC. Extensive experiments show that GreedyViG beats existing ViG, CNN, and ViT architectures in terms of accuracy, GMACs, and parameters on image classification, object detection, instance segmentation, and semantic segmentation tasks. Our smallest model, GreedyViG-S, achieves 81.1% top-1 accuracy on ImageNet-1K, 2.9% higher than Vision GNN and 2.2% higher than Vision HyperGraph Neural Network (ViHGNN), with less GMACs and a similar number of parameters. Our largest model, GreedyViG-B obtains 83.9% top-1 accuracy, 0.2% higher than Vision GNN, with a 66.6% decrease in parameters and a 69% decrease in GMACs. GreedyViG-B also obtains the same accuracy as ViHGNN with a 67.3% decrease in parameters and a 71.3% decrease in GMACs. Our work shows that hybrid CNN-GNN architectures not only provide a new avenue for designing efficient models, but that they can also exceed the performance of current state-of-the-art models.
Mamba, an architecture with RNN-like token mixer of state space model (SSM), was recently introduced to address the quadratic complexity of the attention mechanism and subsequently applied to vision tasks. Nevertheless, the performance of Mamba for vision is often underwhelming when compared with convolutional and attention-based models. In this paper, we delve into the essence of Mamba, and conceptually conclude that Mamba is ideally suited for tasks with long-sequence and autoregressive characteristics. For vision tasks, as image classification does not align with either characteristic, we hypothesize that Mamba is not necessary for this task; Detection and segmentation tasks are also not autoregressive, yet they adhere to the long-sequence characteristic, so we believe it is still worthwhile to explore Mamba's potential for these tasks. To empirically verify our hypotheses, we construct a series of models named MambaOut through stacking Mamba blocks while removing their core token mixer, SSM. Experimental results strongly support our hypotheses. Specifically, our MambaOut model surpasses all visual Mamba models on ImageNet image classification, indicating that Mamba is indeed unnecessary for this task. As for detection and segmentation, MambaOut cannot match the performance of state-of-the-art visual Mamba models, demonstrating the potential of Mamba for long-sequence visual tasks. The code is available at https://github.com/yuweihao/MambaOut
In recent years, convolutional neural networks (CNNs) with channel-wise feature refining mechanisms have brought noticeable benefits to modelling channel dependencies. However, current attention paradigms fail to infer an optimal channel descriptor capable of simultaneously exploiting statistical and spatial relationships among feature maps. In this paper, to overcome this shortcoming, we present a novel channel-wise spatially autocorrelated (CSA) attention mechanism. Inspired by geographical analysis, the proposed CSA exploits the spatial relationships between channels of feature maps to produce an effective channel descriptor. To the best of our knowledge, this is the f irst time that the concept of geographical spatial analysis is utilized in deep CNNs. The proposed CSA imposes negligible learning parameters and light computational overhead to the deep model, making it a powerful yet efficient attention module of choice. We validate the effectiveness of the proposed CSA networks (CSA-Nets) through extensive experiments and analysis on ImageNet, and MS COCO benchmark datasets for image classification, object detection, and instance segmentation. The experimental results demonstrate that CSA-Nets are able to consistently achieve competitive performance and superior generalization than several state-of-the-art attention-based CNNs over different benchmark tasks and datasets.
Recognition of individual components and keypoint detection supported by instance segmentation is crucial to analyze the behavior of agents on the scene. Such systems could be used for surveillance, self-driving cars, and also for medical research, where behavior analysis of laboratory animals is used to confirm the aftereffects of a given medicine. A method capable of solving the aforementioned tasks usually requires a large amount of high-quality hand-annotated data, which takes time and money to produce. In this paper, we propose a method that alleviates the need for manual labeling of laboratory rats. To do so, first, we generate initial annotations with a computer vision-based approach, then through extensive augmentation, we train a deep neural network on the generated data. The final system is capable of instance segmentation, keypoint detection, and body part segmentation even when the objects are heavily occluded.
Data augmentation is a key technique for addressing the challenge of limited datasets, which have become a major component in the training procedures of image processing. Techniques such as geometric transformations and color space adjustments have been thoroughly tested for their ability to artificially expand training datasets and generate semi-realistic data for training purposes. Data augmentation is the most important key to addressing the challenge of limited datasets, which have become a major component of image processing training procedures. Data augmentation techniques, such as geometric transformations and color space adjustments, are thoroughly tested for their ability to artificially expand training datasets and generate semi-realistic data for training purposes. Polygons play a crucial role in instance segmentation and have seen a surge in use across advanced models, such as YOLOv8. Despite their growing popularity, the lack of specialized libraries hampers the polygon-augmentation process. This paper introduces a novel solution to this challenge, embodied in the newly developed AugmenTory library. Notably, AugmenTory offers reduced computational demands in both time and space compared to existing methods. Additionally, the library includes a postprocessing thresholding feature. The AugmenTory package is publicly available on GitHub, where interested users can access the source code: https://github.com/Smartory/AugmenTory
Segment Anything Model (SAM) has achieved impressive performance in many computer vision tasks. However, as a large-scale model, the immense memory and computation costs hinder its practical deployment. In this paper, we propose a post-training quantization (PTQ) framework for Segment Anything Model, namely PTQ4SAM. First, we investigate the inherent bottleneck of SAM quantization attributed to the bimodal distribution in post-Key-Linear activations. We analyze its characteristics from both per-tensor and per-channel perspectives, and propose a Bimodal Integration strategy, which utilizes a mathematically equivalent sign operation to transform the bimodal distribution into a relatively easy-quantized normal distribution offline. Second, SAM encompasses diverse attention mechanisms (i.e., self-attention and two-way cross-attention), resulting in substantial variations in the post-Softmax distributions. Therefore, we introduce an Adaptive Granularity Quantization for Softmax through searching the optimal power-of-two base, which is hardware-friendly. Extensive experimental results across various vision tasks (instance segmentation, semantic segmentation and object detection), datasets and model variants show the superiority of PTQ4SAM. For example, when quantizing SAM-L to 6-bit, we achieve lossless accuracy for instance segmentation, about 0.5\% drop with theoretical 3.9$\times$ acceleration. The code is available at \url{https://github.com/chengtao-lv/PTQ4SAM}.
The segmentation of individual trees from forest point clouds is a crucial task for downstream analyses such as carbon sequestration estimation. Recently, deep-learning-based methods have been proposed which show the potential of learning to segment trees. Since these methods are trained in a supervised way, the question arises how general models can be obtained that are applicable across a wide range of settings. So far, training has been mainly conducted with data from one specific laser scanning type and for specific types of forests. In this work, we train one segmentation model under various conditions, using seven diverse datasets found in literature, to gain insights into the generalization capabilities under domain-shift. Our results suggest that a generalization from coniferous dominated sparse point clouds to deciduous dominated high-resolution point clouds is possible. Conversely, qualitative evidence suggests that generalization from high-resolution to low-resolution point clouds is challenging. This emphasizes the need for forest point clouds with diverse data characteristics for model development. To enrich the available data basis, labeled trees from two previous works were propagated to the complete forest point cloud and are made publicly available at https://doi.org/10.25625/QUTUWU.
Amodal instance segmentation (AIS) predicts the complete shape of the occluded object, including both visible and occluded regions. Because visual clues are lacking, the occluded region is difficult to segment accurately. In human amodal perception, shape-prior knowledge is helpful for AIS. The previous method uses a 2D shape prior by rote memorizing , establishing a shape dictionary and retrieving the closest mask to the segmentation result. However, this approach cannot obtain the shape prior, which is not prestored in the shape dictionary. In this article, to improve generalization ability, we propose a generative invariant shape-prior network (GIN), simulating the human perception process that learns the basic shape, which is invariant to transformations, including translation, rotation, and scaling. We design a novel framework that decouples the learning of shape priors from transformation . GIN is end-to-end trainable and needs no dictionary establishment, making the whole pipeline efficient. GIN outperforms state-of-the-art methods on three public datasets (D2SA, COCOA-cls, and KINS) with large margins.
The Amodal Instance Segmentation (AIS) task aims to infer the complete mask of occluded instance. Under many circumstances, existing methods treat occluded objects as unoccluded ones, and vice versa, leading to inaccurate predictions. This is because existing AIS methods do not explicitly utilize the occlusion rates of each object as supervision. However, occlusion information is critical for the methods to recognize whether the target objects are occluded. Hence we believe it is vital for the method to be distinguishable about the degree of occlusion for each instance. In this paper, a simple yet effective Occlusion-aware transformer-based model, OAFormer, is proposed for accurate amodal instance segmentation. The goal of OAFormer is to learn the occlusion discriminative features. Novel components are proposed to enable OAFormer to be occlusion distinguishable. We conduct extensive experiments on two challenging AIS datasets to evaluate the effectiveness of our method. OAFormer outperforms state-of-the-art methods by large margins.
Instance perception tasks (object detection, instance segmentation, pose estimation, counting) play a key role in industrial applications of visual models. As supervised learning methods suffer from high labeling cost, few-shot learning methods which effectively learn from a limited number of labeled examples are desired. Existing few-shot learning methods primarily focus on a restricted set of tasks, presumably due to the challenges involved in designing a generic model capable of representing diverse tasks in a unified manner. In this paper, we propose UniFS, a universal few-shot instance perception model that unifies a wide range of instance perception tasks by reformulating them into a dynamic point representation learning framework. Additionally, we propose Structure-Aware Point Learning (SAPL) to exploit the higher-order structural relationship among points to further enhance representation learning. Our approach makes minimal assumptions about the tasks, yet it achieves competitive results compared to highly specialized and well optimized specialist models. Codes and data are available at https://github.com/jin-s13/UniFS.
This paper introduces YOLOv8-TO, a novel approach for reverse engineering of topology-optimized structures into interpretable geometric parameters using the YOLOv8 instance segmentation model. Density-based topology optimization methods require post-processing to convert the optimal density distribution into a parametric representation for design exploration and integration with CAD tools. Traditional methods such as skeletonization struggle with complex geometries and require manual intervention. YOLOv8-TO addresses these challenges by training a custom YOLOv8 model to automatically detect and reconstruct structural components from binary density distributions. The model is trained on a diverse dataset of both optimized and random structures generated using the Moving Morphable Components method. A custom reconstruction loss function based on the dice coefficient of the predicted geometry is used to train the new regression head of the model via self-supervised learning. The method is evaluated on test sets generated from different topology optimization methods, including out-of-distribution samples, and compared against a skeletonization approach. Results show that YOLOv8-TO significantly outperforms skeletonization in reconstructing visually and structurally similar designs. The method showcases an average improvement of 13.84% in the Dice coefficient, with peak enhancements reaching 20.78%. The method demonstrates good generalization to complex geometries and fast inference times, making it suitable for integration into design workflows using regular workstations. Limitations include the sensitivity to non-max suppression thresholds. YOLOv8-TO represents a significant advancement in topology optimization post-processing, enabling efficient and accurate reverse engineering of optimized structures for design exploration and manufacturing.
Current state-of-the-art two-stage models on instance segmentation task suffer from several types of imbalances. In this paper, we address the Intersection over the Union (IoU) distribution imbalance of positive input Regions of Interest (RoIs) during the training of the second stage. Our Self-Balanced R-CNN (SBR-CNN), an evolved version of the Hybrid Task Cascade (HTC) model, brings brand new loop mechanisms of bounding box and mask refinements. With an improved Generic RoI Extraction (GRoIE), we also address the feature-level imbalance at the Feature Pyramid Network (FPN) level, originated by a non-uniform integration between low- and high-level features from the backbone layers. In addition, the redesign of the architecture heads toward a fully convolutional approach with FCC further reduces the number of parameters and obtains more clues to the connection between the task to solve and the layers used. Moreover, our SBR-CNN model shows the same or even better improvements if adopted in conjunction with other state-of-the-art models. In fact, with a lightweight ResNet-50 as backbone, evaluated on COCO minival 2017 dataset, our model reaches 45.3% and 41.5% AP for object detection and instance segmentation, with 12 epochs and without extra tricks. The code is available at https://github.com/IMPLabUniPr/mmdetection/tree/sbr_cnn
Recently, event-based vision sensors have gained attention for autonomous driving applications, as conventional RGB cameras face limitations in handling challenging dynamic conditions. However, the availability of real-world and synthetic event-based vision datasets remains limited. In response to this gap, we present SEVD, a first-of-its-kind multi-view ego, and fixed perception synthetic event-based dataset using multiple dynamic vision sensors within the CARLA simulator. Data sequences are recorded across diverse lighting (noon, nighttime, twilight) and weather conditions (clear, cloudy, wet, rainy, foggy) with domain shifts (discrete and continuous). SEVD spans urban, suburban, rural, and highway scenes featuring various classes of objects (car, truck, van, bicycle, motorcycle, and pedestrian). Alongside event data, SEVD includes RGB imagery, depth maps, optical flow, semantic, and instance segmentation, facilitating a comprehensive understanding of the scene. Furthermore, we evaluate the dataset using state-of-the-art event-based (RED, RVT) and frame-based (YOLOv8) methods for traffic participant detection tasks and provide baseline benchmarks for assessment. Additionally, we conduct experiments to assess the synthetic event-based dataset's generalization capabilities. The dataset is available at https://eventbasedvision.github.io/SEVD
Object detection is a mature problem in autonomous driving with pedestrian detection being one of the first deployed algorithms. It has been comprehensively studied in the literature. However, object detection is relatively less explored for fisheye cameras used for surround-view near field sensing. The standard bounding box representation fails in fisheye cameras due to heavy radial distortion, particularly in the periphery. To mitigate this, we explore extending the standard object detection output representation of bounding box. We design rotated bounding boxes, ellipse, generic polygon as polar arc/angle representations and define an instance segmentation mIOU metric to analyze these representations. The proposed model FisheyeDetNet with polygon outperforms others and achieves a mAP score of 49.5 % on Valeo fisheye surround-view dataset for automated driving applications. This dataset has 60K images captured from 4 surround-view cameras across Europe, North America and Asia. To the best of our knowledge, this is the first detailed study on object detection on fisheye cameras for autonomous driving scenarios.