The referring video object segmentation task (RVOS) aims to segment object instances in a given video referred by a language expression in all video frames. Due to the requirement of understanding cross-modal semantics within individual instances, this task is more challenging than the traditional semi-supervised video object segmentation where the ground truth object masks in the first frame are given. With the great achievement of Transformer in object detection and object segmentation, RVOS has been made remarkable progress where ReferFormer achieved the state-of-the-art performance. In this work, based on the strong baseline framework--ReferFormer, we propose several tricks to boost further, including cyclical learning rates, semi-supervised approach, and test-time augmentation inference. The improved ReferFormer ranks 2nd place on CVPR2022 Referring Youtube-VOS Challenge.
We analyzed the network structure of real-time object detection models and found that the features in the feature concatenation stage are very rich. Applying an attention module here can effectively improve the detection accuracy of the model. However, the commonly used attention module or self-attention module shows poor performance in detection accuracy and inference efficiency. Therefore, we propose a novel self-attention module, called 2D local feature superimposed self-attention, for the feature concatenation stage of the neck network. This self-attention module reflects global features through local features and local receptive fields. We also propose and optimize an efficient decoupled head and AB-OTA, and achieve SOTA results. Average precisions of 49.0% (71FPS, 14ms), 46.1% (85FPS, 11.7ms), and 39.1% (107FPS, 9.3ms) were obtained for large, medium, and small-scale models built using our proposed improvements. Our models exceeded YOLOv5 by 0.8% -- 3.1% in average precision.
Region extraction is necessary in a wide range of applications, from object detection in autonomous driving to analysis of subcellular morphology in cell biology. There exist two main approaches: convex hull extraction, for which exact and efficient algorithms exist and concave hulls, which are better at capturing real-world shapes but do not have a single solution. Especially in the context of a uniform grid, concave hull algorithms are largely approximate, sacrificing region integrity for spatial and temporal efficiency. In this study, we present a novel algorithm that can provide vertex-minimized concave hulls with maximal (i.e. pixel-perfect) resolution and is tunable for speed-efficiency tradeoffs. Our method provides advantages in multiple downstream applications including data compression, retrieval, visualization, and analysis. To demonstrate the practical utility of our approach, we focus on image compression. We demonstrate significant improvements through context-dependent compression on disparate regions within a single image (entropy encoding for noisy and predictive encoding for the structured regions). We show that these improvements range from biomedical images to natural images. Beyond image compression, our algorithm can be applied more broadly to aid in a wide range of practical applications for data retrieval, visualization, and analysis.
For nearly a decade, the COCO dataset has been the central test bed of research in object detection. According to the recent benchmarks, however, it seems that performance on this dataset has started to saturate. One possible reason can be that perhaps it is not large enough for training deep models. To address this limitation, here we introduce two complementary datasets to COCO: i) COCO_OI, composed of images from COCO and OpenImages (from their 80 classes in common) with 1,418,978 training bounding boxes over 380,111 images, and 41,893 validation bounding boxes over 18,299 images, and ii) ObjectNet_D containing objects in daily life situations (originally created for object recognition known as ObjectNet; 29 categories in common with COCO). The latter can be used to test the generalization ability of object detectors. We evaluate some models on these datasets and pinpoint the source of errors. We encourage the community to utilize these datasets for training and testing object detection models. Code and data is available at https://github.com/aliborji/COCO_OI.
We endeavor on a rarely explored task named Insubstantial Object Detection (IOD), which aims to localize the object with following characteristics: (1) amorphous shape with indistinct boundary; (2) similarity to surroundings; (3) absence in color. Accordingly, it is far more challenging to distinguish insubstantial objects in a single static frame and the collaborative representation of spatial and temporal information is crucial. Thus, we construct an IOD-Video dataset comprised of 600 videos (141,017 frames) covering various distances, sizes, visibility, and scenes captured by different spectral ranges. In addition, we develop a spatio-temporal aggregation framework for IOD, in which different backbones are deployed and a spatio-temporal aggregation loss (STAloss) is elaborately designed to leverage the consistency along the time axis. Experiments conducted on IOD-Video dataset demonstrate that spatio-temporal aggregation can significantly improve the performance of IOD. We hope our work will attract further researches into this valuable yet challenging task. The code will be available at: \url{https://github.com/CalayZhou/IOD-Video}.
This technical report presents the 1st place winning solution for the Waymo Open Dataset 3D semantic segmentation challenge 2022. Our network, termed LidarMultiNet, unifies the major LiDAR perception tasks such as 3D semantic segmentation, object detection, and panoptic segmentation in a single framework. At the core of LidarMultiNet is a strong 3D voxel-based encoder-decoder network with a novel Global Context Pooling (GCP) module extracting global contextual features from a LiDAR frame to complement its local features. An optional second stage is proposed to refine the first-stage segmentation or generate accurate panoptic segmentation results. Our solution achieves a mIoU of 71.13 and is the best for most of the 22 classes on the Waymo 3D semantic segmentation test set, outperforming all the other 3D semantic segmentation methods on the official leaderboard. We demonstrate for the first time that major LiDAR perception tasks can be unified in a single strong network that can be trained end-to-end.
Open-vocabulary object detection (OVD) aims to scale up vocabulary size to detect objects of novel categories beyond the training vocabulary. Recent work resorts to the rich knowledge in pre-trained vision-language models. However, existing methods are ineffective in proposal-level vision-language alignment. Meanwhile, the models usually suffer from confidence bias toward base categories and perform worse on novel ones. To overcome the challenges, we present MEDet, a novel and effective OVD framework with proposal mining and prediction equalization. First, we design an online proposal mining to refine the inherited vision-semantic knowledge from coarse to fine, allowing for proposal-level detection-oriented feature alignment. Second, based on causal inference theory, we introduce a class-wise backdoor adjustment to reinforce the predictions on novel categories to improve the overall OVD performance. Extensive experiments on COCO and LVIS benchmarks verify the superiority of MEDet over the competing approaches in detecting objects of novel categories, e.g., 32.6% AP50 on COCO and 22.4% mask mAP on LVIS.
Most black-box adversarial attack schemes for object detectors mainly face two shortcomings: requiring access to the target model and generating inefficient adversarial examples (failing to make objects disappear in large numbers). To overcome these shortcomings, we propose a black-box adversarial attack scheme based on semantic segmentation and model inversion (SSMI). We first locate the position of the target object using semantic segmentation techniques. Next, we design a neighborhood background pixel replacement to replace the target region pixels with background pixels to ensure that the pixel modifications are not easily detected by human vision. Finally, we reconstruct a machine-recognizable example and use the mask matrix to select pixels in the reconstructed example to modify the benign image to generate an adversarial example. Detailed experimental results show that SSMI can generate efficient adversarial examples to evade human-eye perception and make objects of interest disappear. And more importantly, SSMI outperforms existing same kinds of attacks. The maximum increase in new and disappearing labels is 16%, and the maximum decrease in mAP metrics for object detection is 36%.
While behavior learning has made impressive progress in recent times, it lags behind computer vision and natural language processing due to its inability to leverage large, human-generated datasets. Human behaviors have wide variance, multiple modes, and human demonstrations typically do not come with reward labels. These properties limit the applicability of current methods in Offline RL and Behavioral Cloning to learn from large, pre-collected datasets. In this work, we present Behavior Transformer (BeT), a new technique to model unlabeled demonstration data with multiple modes. BeT retrofits standard transformer architectures with action discretization coupled with a multi-task action correction inspired by offset prediction in object detection. This allows us to leverage the multi-modal modeling ability of modern transformers to predict multi-modal continuous actions. We experimentally evaluate BeT on a variety of robotic manipulation and self-driving behavior datasets. We show that BeT significantly improves over prior state-of-the-art work on solving demonstrated tasks while capturing the major modes present in the pre-collected datasets. Finally, through an extensive ablation study, we analyze the importance of every crucial component in BeT. Videos of behavior generated by BeT are available at https://notmahi.github.io/bet
The step function is one of the simplest and most natural activation functions for deep neural networks (DNNs). As it counts 1 for positive variables and 0 for others, its intrinsic characteristics (e.g., discontinuity and no viable information of subgradients) impede its development for several decades. Even if there is an impressive body of work on designing DNNs with continuous activation functions that can be deemed as surrogates of the step function, it is still in the possession of some advantageous properties, such as complete robustness to outliers and being capable of attaining the best learning-theoretic guarantee of predictive accuracy. Hence, in this paper, we aim to train DNNs with the step function used as an activation function (dubbed as 0/1 DNNs). We first reformulate 0/1 DNNs as an unconstrained optimization problem and then solve it by a block coordinate descend (BCD) method. Moreover, we acquire closed-form solutions for sub-problems of BCD as well as its convergence properties. Furthermore, we also integrate $\ell_{2,0}$-regularization into 0/1 DNN to accelerate the training process and compress the network scale. As a result, the proposed algorithm has a high performance on classifying MNIST and Fashion-MNIST datasets. As a result, the proposed algorithm has a desirable performance on classifying MNIST, FashionMNIST, Cifar10, and Cifar100 datasets.
In the pursuit of achieving ever-increasing accuracy, large and complex neural networks are usually developed. Such models demand high computational resources and therefore cannot be deployed on edge devices. It is of great interest to build resource-efficient general purpose networks due to their usefulness in several application areas. In this work, we strive to effectively combine the strengths of both CNN and Transformer models and propose a new efficient hybrid architecture EdgeNeXt. Specifically in EdgeNeXt, we introduce split depth-wise transpose attention (STDA) encoder that splits input tensors into multiple channel groups and utilizes depth-wise convolution along with self-attention across channel dimensions to implicitly increase the receptive field and encode multi-scale features. Our extensive experiments on classification, detection and segmentation tasks, reveal the merits of the proposed approach, outperforming state-of-the-art methods with comparatively lower compute requirements. Our EdgeNeXt model with 1.3M parameters achieves 71.2% top-1 accuracy on ImageNet-1K, outperforming MobileViT with an absolute gain of 2.2% with 28% reduction in FLOPs. Further, our EdgeNeXt model with 5.6M parameters achieves 79.4% top-1 accuracy on ImageNet-1K. The code and models are available at https://t.ly/_Vu9.
Recent advance in 2D CNNs has revealed that large kernels are important. However, when directly applying large convolutional kernels in 3D CNNs, severe difficulties are met, where those successful module designs in 2D become surprisingly ineffective on 3D networks, including the popular depth-wise convolution. To address this vital challenge, we instead propose the spatial-wise partition convolution and its large-kernel module. As a result, it avoids the optimization and efficiency issues of naive 3D large kernels. Our large-kernel 3D CNN network, LargeKernel3D, yields notable improvement in 3D tasks of semantic segmentation and object detection. It achieves 73.9% mIoU on the ScanNetv2 semantic segmentation and 72.8% NDS nuScenes object detection benchmarks, ranking 1st on the nuScenes LIDAR leaderboard. The performance further boosts to 74.2% NDS with a simple multi-modal fusion. In addition, LargeKernel3D can be scaled to 17x17x17 kernel size on Waymo 3D object detection. For the first time, we show that large kernels are feasible and essential for 3D visual tasks.
Based on multiple instance detection networks (MIDN), plenty of works have contributed tremendous efforts to weakly supervised object detection (WSOD). However, most methods neglect the fact that the overwhelming negative instances exist in each image during the training phase, which would mislead the training and make the network fall into local minima. To tackle this problem, an online progressive instance-balanced sampling (OPIS) algorithm based on hard sampling and soft sampling is proposed in this paper. The algorithm includes two modules: a progressive instance balance (PIB) module and a progressive instance reweighting (PIR) module. The PIB module combining random sampling and IoU-balanced sampling progressively mines hard negative instances while balancing positive instances and negative instances. The PIR module further utilizes classifier scores and IoUs of adjacent refinements to reweight the weights of positive instances for making the network focus on positive instances. Extensive experimental results on the PASCAL VOC 2007 and 2012 datasets demonstrate the proposed method can significantly improve the baseline, which is also comparable to many existing state-of-the-art results. In addition, compared to the baseline, the proposed method requires no extra network parameters and the supplementary training overheads are small, which could be easily integrated into other methods based on the instance classifier refinement paradigm.
The Instance Segmentation task, an extension of the well-known Object Detection task, is of great help in many areas, such as precision agriculture: being able to automatically identify plant organs and the possible diseases associated with them, allows to effectively scale and automate crop monitoring and its diseases control. To address the problem related to early disease detection and diagnosis on vines plants, a new dataset has been created with the goal of advancing the state-of-the-art of diseases recognition via instance segmentation approaches. This was achieved by gathering images of leaves and clusters of grapes affected by diseases in their natural context. The dataset contains photos of 10 object types which include leaves and grapes with and without symptoms of the eight more common grape diseases, with a total of 17,706 labeled instances in 1,092 images. Multiple statistical measures are proposed in order to offer a complete view on the characteristics of the dataset. Preliminary results for the object detection and instance segmentation tasks reached by the models Mask R-CNN and R^3-CNN are provided as baseline, demonstrating that the procedure is able to reach promising results about the objective of automatic diseases' symptoms recognition.
Nowadays, Semi-Supervised Object Detection (SSOD) is a hot topic, since, while it is rather easy to collect images for creating a new dataset, labeling them is still an expensive and time-consuming task. One of the successful methods to take advantage of raw images on a Semi-Supervised Learning (SSL) setting is the Mean Teacher technique, where the operations of pseudo-labeling by the Teacher and the Knowledge Transfer from the Student to the Teacher take place simultaneously. However, the pseudo-labeling by thresholding is not the best solution since the confidence value is not strictly related to the prediction uncertainty, not permitting to safely filter predictions. In this paper, we introduce an additional classification task for bounding box localization to improve the filtering of the predicted bounding boxes and obtain higher quality on Student training. Furthermore, we empirically prove that bounding box regression on the unsupervised part can equally contribute to the training as much as category classification. Our experiments show that our IL-net (Improving Localization net) increases SSOD performance by 1.14% AP on COCO dataset in limited-annotation regime. The code is available at https://github.com/IMPLabUniPr/unbiased-teacher/tree/ilnet
Object detection is a fundamental vision task. It has been highly researched in academia and has been widely adopted in industry. Average Precision (AP) is the standard score for evaluating object detectors. Our understanding of the subtleties of this score, however, is limited. Here, we quantify the sensitivity of AP to bounding box perturbations and show that AP is very sensitive to small translations. Only one pixel shift is enough to drop the mAP of a model by 8.4%. The mAP drop over small objects with only one pixel shift is 23.1%. The corresponding numbers when ground-truth (GT) boxes are used as predictions are 23% and 41.7%, respectively. These results explain why achieving higher mAP becomes increasingly harder as models get better. We also investigate the effect of box scaling on AP. Code and data is available at https://github.com/aliborji/AP_Box_Perturbation.
In this research, we propose a new 3D object detector with a trustworthy depth estimation, dubbed BEVDepth, for camera-based Bird's-Eye-View (BEV) 3D object detection. Our work is based on a key observation -- depth estimation in recent approaches is surprisingly inadequate given the fact that depth is essential to camera 3D detection. Our BEVDepth resolves this by leveraging explicit depth supervision. A camera-awareness depth estimation module is also introduced to facilitate the depth predicting capability. Besides, we design a novel Depth Refinement Module to counter the side effects carried by imprecise feature unprojection. Aided by customized Efficient Voxel Pooling and multi-frame mechanism, BEVDepth achieves the new state-of-the-art 60.9% NDS on the challenging nuScenes test set while maintaining high efficiency. For the first time, the NDS score of a camera model reaches 60%.
Current perception models in autonomous driving heavily rely on large-scale labelled 3D data, which is both costly and time-consuming to annotate. This work proposes a solution to reduce the dependence on labelled 3D training data by leveraging pre-training on large-scale unlabeled outdoor LiDAR point clouds using masked autoencoders (MAE). While existing masked point autoencoding methods mainly focus on small-scale indoor point clouds or pillar-based large-scale outdoor LiDAR data, our approach introduces a new self-supervised masked occupancy pre-training method called Occupancy-MAE, specifically designed for voxel-based large-scale outdoor LiDAR point clouds. Occupancy-MAE takes advantage of the gradually sparse voxel occupancy structure of outdoor LiDAR point clouds and incorporates a range-aware random masking strategy and a pretext task of occupancy prediction. By randomly masking voxels based on their distance to the LiDAR and predicting the masked occupancy structure of the entire 3D surrounding scene, Occupancy-MAE encourages the extraction of high-level semantic information to reconstruct the masked voxel using only a small number of visible voxels. Extensive experiments demonstrate the effectiveness of Occupancy-MAE across several downstream tasks. For 3D object detection, Occupancy-MAE reduces the labelled data required for car detection on the KITTI dataset by half and improves small object detection by approximately 2% in AP on the Waymo dataset. For 3D semantic segmentation, Occupancy-MAE outperforms training from scratch by around 2% in mIoU. For multi-object tracking, Occupancy-MAE enhances training from scratch by approximately 1% in terms of AMOTA and AMOTP. Codes are publicly available at https://github.com/chaytonmin/Occupancy-MAE.
Over the years, datasets have been developed for various object detection tasks. Object detection in the maritime domain is essential for the safety and navigation of ships. However, there is still a lack of publicly available large-scale datasets in the maritime domain. To overcome this challenge, we present KOLOMVERSE, an open large-scale image dataset for object detection in the maritime domain by KRISO (Korea Research Institute of Ships and Ocean Engineering). We collected 5,845 hours of video data captured from 21 territorial waters of South Korea. Through an elaborate data quality assessment process, we gathered around 2,151,470 4K resolution images from the video data. This dataset considers various environments: weather, time, illumination, occlusion, viewpoint, background, wind speed, and visibility. The KOLOMVERSE consists of five classes (ship, buoy, fishnet buoy, lighthouse and wind farm) for maritime object detection. The dataset has images of 3840$\times$2160 pixels and to our knowledge, it is by far the largest publicly available dataset for object detection in the maritime domain. We performed object detection experiments and evaluated our dataset on several pre-trained state-of-the-art architectures to show the effectiveness and usefulness of our dataset. The dataset is available at: \url{https://github.com/MaritimeDataset/KOLOMVERSE}.
Deep convolutional neural network with increased number of parameters has achieved improved precision in task of object detection on natural images, where objects of interests are annotated with horizontal boundary boxes. On aerial images captured from the bird-view perspective, these improvements on model architecture and deeper convolutional layers can also boost the performance on oriented object detection task. However, it is hard to directly apply those state-of-the-art object detectors on the devices with limited computation resources, which necessitates lightweight models through model compression. In order to address this issue, we present a model compression method for rotated object detection on aerial images by knowledge distillation, namely KD-RNet. With a well-trained teacher oriented object detector with a large number of parameters, the obtained object category and location information are both transferred to a compact student network in KD-RNet by collaborative training strategy. Transferring the category information is achieved by knowledge distillation on predicted probability distribution, and a soft regression loss is adopted for handling displacement in location information transfer. The experimental result on a large-scale aerial object detection dataset (DOTA) demonstrates that the proposed KD-RNet model can achieve improved mean-average precision (mAP) with reduced number of parameters, at the same time, KD-RNet boost the performance on providing high quality detections with higher overlap with groundtruth annotations.