In this report, we present PP-YOLOE, an industrial state-of-the-art object detector with high performance and friendly deployment. We optimize on the basis of the previous PP-YOLOv2, using anchor-free paradigm, more powerful backbone and neck equipped with CSPRepResStage, ET-head and dynamic label assignment algorithm TAL. We provide s/m/l/x models for different practice scenarios. As a result, PP-YOLOE-l achieves 51.4 mAP on COCO test-dev and 78.1 FPS on Tesla V100, yielding a remarkable improvement of (+1.9 AP, +13.35% speed up) and (+1.3 AP, +24.96% speed up), compared to the previous state-of-the-art industrial models PP-YOLOv2 and YOLOX respectively. Further, PP-YOLOE inference speed achieves 149.2 FPS with TensorRT and FP16-precision. We also conduct extensive experiments to verify the effectiveness of our designs. Source code and pre-trained models are available at https://github.com/PaddlePaddle/PaddleDetection.
This study addresses the issue of fusing infrared and visible images that appear differently for object detection. Aiming at generating an image of high visual quality, previous approaches discover commons underlying the two modalities and fuse upon the common space either by iterative optimization or deep networks. These approaches neglect that modality differences implying the complementary information are extremely important for both fusion and subsequent detection task. This paper proposes a bilevel optimization formulation for the joint problem of fusion and detection, and then unrolls to a target-aware Dual Adversarial Learning (TarDAL) network for fusion and a commonly used detection network. The fusion network with one generator and dual discriminators seeks commons while learning from differences, which preserves structural information of targets from the infrared and textural details from the visible. Furthermore, we build a synchronized imaging system with calibrated infrared and optical sensors, and collect currently the most comprehensive benchmark covering a wide range of scenarios. Extensive experiments on several public datasets and our benchmark demonstrate that our method outputs not only visually appealing fusion but also higher detection mAP than the state-of-the-art approaches.
Object detection, for the most part, has been formulated in the euclidean space, where euclidean or spherical geodesic distances measure the similarity of an image region to an object class prototype. In this work, we study whether a hyperbolic geometry better matches the underlying structure of the object classification space. We incorporate a hyperbolic classifier in two-stage, keypoint-based, and transformer-based object detection architectures and evaluate them on large-scale, long-tailed, and zero-shot object detection benchmarks. In our extensive experimental evaluations, we observe categorical class hierarchies emerging in the structure of the classification space, resulting in lower classification errors and boosting the overall object detection performance.
Integrating different representations from complementary sensing modalities is crucial for robust scene interpretation in autonomous driving. While deep learning architectures that fuse vision and range data for 2D object detection have thrived in recent years, the corresponding modalities can degrade in adverse weather or lighting conditions, ultimately leading to a drop in performance. Although domain adaptation methods attempt to bridge the domain gap between source and target domains, they do not readily extend to heterogeneous data distributions. In this work, we propose an unsupervised domain adaptation framework, which adapts a 2D object detector for RGB and lidar sensors to one or more target domains featuring adverse weather conditions. Our proposed approach consists of three components. First, a data augmentation scheme that simulates weather distortions is devised to add domain confusion and prevent overfitting on the source data. Second, to promote cross-domain foreground object alignment, we leverage the complementary features of multiple modalities through a multi-scale entropy-weighted domain discriminator. Finally, we use carefully designed pretext tasks to learn a more robust representation of the target domain data. Experiments performed on the DENSE dataset show that our method can substantially alleviate the domain gap under the single-target domain adaptation (STDA) setting and the less explored yet more general multi-target domain adaptation (MTDA) setting.
In this paper, we consider the imperfection within machine learning-based 2D object detection and its impact on safety. We address a special sub-type of performance limitations: the prediction bounding box cannot be perfectly aligned with the ground truth, but the computed Intersection-over-Union metric is always larger than a given threshold. Under such type of performance limitation, we formally prove the minimum required bounding box enlargement factor to cover the ground truth. We then demonstrate that the factor can be mathematically adjusted to a smaller value, provided that the motion planner takes a fixed-length buffer in making its decisions. Finally, observing the difference between an empirically measured enlargement factor and our formally derived worst-case enlargement factor offers an interesting connection between the quantitative evidence (demonstrated by statistics) and the qualitative evidence (demonstrated by worst-case analysis).
Object detection serves as a significant step in improving performance of complex downstream computer vision tasks. It has been extensively studied for many years now and current state-of-the-art 2D object detection techniques proffer superlative results even in complex images. In this chapter, we discuss the geometry-based pioneering works in object detection, followed by the recent breakthroughs that employ deep learning. Some of these use a monolithic architecture that takes a RGB image as input and passes it to a feed-forward ConvNet or vision Transformer. These methods, thereby predict class-probability and bounding-box coordinates, all in a single unified pipeline. Two-stage architectures on the other hand, first generate region proposals and then feed it to a CNN to extract features and predict object category and bounding-box. We also elaborate upon the applications of object detection in video event recognition, to achieve better fine-grained video classification performance. Further, we highlight recent datasets for 2D object detection both in images and videos, and present a comparative performance summary of various state-of-the-art object detection techniques.
Accurate 3D object detection is a key part of the perception module for autonomous vehicles. A better understanding of the objects in 3D facilitates better decision-making and path planning. RGB Cameras and LiDAR are the most commonly used sensors in autonomous vehicles for environment perception. Many approaches have shown promising results for 2D detection with RGB Images, but efficiently localizing small objects like pedestrians in the 3D point cloud of large scenes has remained a challenging area of research. We propose a novel method, Frustum-PointPillars, for 3D object detection using LiDAR data. Instead of solely relying on point cloud features, we leverage the mature field of 2D object detection to reduce the search space in the 3D space. Then, we use the Pillar Feature Encoding network for object localization in the reduced point cloud. We also propose a novel approach for masking point clouds to further improve the localization of objects. We train our network on the KITTI dataset and perform experiments to show the effectiveness of our network. On the KITTI test set our method outperforms other multi-sensor SOTA approaches for 3D pedestrian localization (Bird’s Eye View) while achieving a significantly faster runtime of 14 Hz.
We present in this paper a novel query formulation using dynamic anchor boxes for DETR (DEtection TRansformer) and offer a deeper understanding of the role of queries in DETR. This new formulation directly uses box coordinates as queries in Transformer decoders and dynamically updates them layer-by-layer. Using box coordinates not only helps using explicit positional priors to improve the query-to-feature similarity and eliminate the slow training convergence issue in DETR, but also allows us to modulate the positional attention map using the box width and height information. Such a design makes it clear that queries in DETR can be implemented as performing soft ROI pooling layer-by-layer in a cascade manner. As a result, it leads to the best performance on MS-COCO benchmark among the DETR-like detection models under the same setting, e.g., AP 45.7\% using ResNet50-DC5 as backbone trained in 50 epochs. We also conducted extensive experiments to confirm our analysis and verify the effectiveness of our methods. Code is available at \url{https://github.com/SlongLiu/DAB-DETR}.
Mobile headsets should be capable of understanding 3D physical environments to offer a truly immersive experience for augmented/mixed reality (AR/MR). However, their small form-factor and limited computation resources make it extremely challenging to execute in real-time 3D vision algorithms, which are known to be more compute-intensive than their 2D counterparts. In this paper, we propose DeepMix, a mobility-aware, lightweight, and hybrid 3D object detection framework for improving the user experience of AR/MR on mobile headsets. Motivated by our analysis and evaluation of state-of-the-art 3D object detection models, DeepMix intelligently combines edge-assisted 2D object detection and novel, on-device 3D bounding box estimations that leverage depth data captured by headsets. This leads to low end-to-end latency and significantly boosts detection accuracy in mobile scenarios. A unique feature of DeepMix is that it fully exploits the mobility of headsets to fine-tune detection results and boost detection accuracy. To the best of our knowledge, DeepMix is the first 3D object detection that achieves 30 FPS (an end-to-end latency much lower than the 100 ms stringent requirement of interactive AR/MR). We implement a prototype of DeepMix on Microsoft HoloLens and evaluate its performance via both extensive controlled experiments and a user study with 30+ participants. DeepMix not only improves detection accuracy by 9.1--37.3% but also reduces end-to-end latency by 2.68--9.15x, compared to the baseline that uses existing 3D object detection models.
This paper presents a grounded language-image pre-training (GLIP) model for learning object-level, language-aware, and semantic-rich visual representations. GLIP unifies object detection and phrase grounding for pre-training. The unification brings two benefits: 1) it allows GLIP to learn from both detection and grounding data to improve both tasks and bootstrap a good grounding model; 2) GLIP can leverage massive image-text pairs by generating grounding boxes in a self-training fashion, making the learned representation semantic-rich. In our experiments, we pre-train GLIP on 27M grounding data, including 3M human-annotated and 24M web-crawled image-text pairs. The learned representations demonstrate strong zero-shot and few-shot transferability to various object-level recognition tasks. 1) When directly evaluated on COCO and LVIS (without seeing any images in COCO during pre-training), GLIP achieves 49.8 AP and 26.9 AP, respectively, surpassing many supervised baselines. 2) After fine-tuned on COCO, GLIP achieves 60.8 AP on val and 61.5 AP on test-dev, surpassing prior SoTA. 3) When transferred to 13 downstream object detection tasks, a 1-shot GLIP rivals with a fully-supervised Dynamic Head. Code is released at https://github.com/microsoft/GLIP.
Unlike 2D object detection where all RoI features come from grid pixels, the RoI feature extraction of 3D point cloud object detection is more diverse. In this paper, we first compare and analyze the differences in structure and performance between the two state-of-the-art models PV-RCNN and Voxel-RCNN. Then, we find that the performance gap between the two models does not come from point information, but structural information. The voxel features contain more structural information because they do quantization instead of downsampling to point cloud so that they can contain basically the complete information of the whole point cloud. The stronger structural information in voxel features makes the detector have higher performance in our experiments even if the voxel features don't have accurate location information. Then, we propose that structural information is the key to 3D object detection. Based on the above conclusion, we propose a Self-Attention RoI Feature Extractor (SARFE) to enhance structural information of the feature extracted from 3D proposals. SARFE is a plug-and-play module that can be easily used on existing 3D detectors. Our SARFE is evaluated on both KITTI dataset and Waymo Open dataset. With the newly introduced SARFE, we improve the performance of the state-of-the-art 3D detectors by a large margin in cyclist on KITTI dataset while keeping real-time capability.
Vehicle velocity and inter-vehicle distance estimation are essential for ADAS (Advanced driver-assistance systems) and autonomous vehicles. To save the cost of expensive ranging sensors, recent studies focus on using a low-cost monocular camera to perceive the environment around the vehicle in a data-driven fashion. Existing approaches treat each vehicle independently for perception and cause inconsistent estimation. Furthermore, important information like context and spatial relation in 2D object detection is often neglected in the velocity estimation pipeline. In this paper, we explore the relationship between vehicles of the same frame with a global-relative-constraint (GLC) loss to encourage consistent estimation. A novel multi-stream attention network (MSANet) is proposed to extract different aspects of features, e.g., spatial and contextual features, for joint vehicle velocity and inter-vehicle distance estimation. Experiments show the effectiveness and robustness of our proposed approach. MSANet outperforms state-of-the-art algorithms on both the KITTI dataset and TuSimple velocity dataset.
Visual perception in autonomous driving is a crucial part of a vehicle to navigate safely and sustainably in different traffic conditions. However, in bad weather such as heavy rain and haze, the performance of visual perception is greatly affected by several degrading effects. Recently, deep learning-based perception methods have addressed multiple degrading effects to reflect real-world bad weather cases but have shown limited success due to 1) high computational costs for deployment on mobile devices and 2) poor relevance between image enhancement and visual perception in terms of the model ability. To solve these issues, we propose a task-driven image enhancement network connected to the high-level vision task, which takes in an image corrupted by bad weather as input. Specifically, we introduce a novel low memory network to reduce most of the layer connections of dense blocks for less memory and computational cost while maintaining high performance. We also introduce a new task-driven training strategy to robustly guide the high-level task model suitable for both high-quality restoration of images and highly accurate perception. Experiment results demonstrate that the proposed method improves the performance among lane and 2D object detection, and depth estimation largely under adverse weather in terms of both low memory and accuracy.
3D object detection often involves complicated training and testing pipelines, which require substantial domain knowledge about individual datasets. Inspired by recent non-maximum suppression-free 2D object detection models, we propose a 3D object detection architecture on point clouds. Our method models 3D object detection as message passing on a dynamic graph, generalizing the DGCNN framework to predict a set of objects. In our construction, we remove the necessity of post-processing via object confidence aggregation or non-maximum suppression. To facilitate object detection from sparse point clouds, we also propose a set-to-set distillation approach customized to 3D detection. This approach aligns the outputs of the teacher model and the student model in a permutation-invariant fashion, significantly simplifying knowledge distillation for the 3D detection task. Our method achieves state-of-the-art performance on autonomous driving benchmarks. We also provide abundant analysis of the detection model and distillation framework.
Prior work on 6-DoF object pose estimation has largely focused on instance-level processing, in which a textured CAD model is available for each object being detected. Category-level 6-DoF pose estimation represents an important step toward developing robotic vision systems that operate in unstructured, real-world scenarios. In this work, we propose a single-stage, keypoint-based approach for category-level object pose estimation that operates on unknown object instances within a known category using a single RGB image as input. The proposed network performs 2D object detection, detects 2D keypoints, estimates 6-DoF pose, and regresses relative bounding cuboid dimensions. These quantities are estimated in a sequential fashion, leveraging the recent idea of convGRU for propagating information from easier tasks to those that are more difficult. We favor simplicity in our design choices: generic cuboid vertex coordinates, single-stage network, and monocular RGB input. We conduct extensive experiments on the challenging Objectron benchmark, outperforming state-of-the-art methods on the 3D IoU metric (27.6% higher than the MobilePose single-stage approach and 7.1% higher than the related two-stage approach).
Deep neural networks have proven increasingly important for automotive scene understanding with new algorithms offering constant improvements of the detection performance. However, there is little emphasis on experiences and needs for deployment in embedded environments. We therefore perform a case study of the deployment of two representative object detection networks on an edge AI platform. In particular, we consider RetinaNet for image-based 2D object detection and PointPillars for LiDAR-based 3D object detection. We describe the modifications necessary to convert the algorithms from a PyTorch training environment to the deployment environment taking into account the available tools. We evaluate the runtime of the deployed DNN using two different libraries, TensorRT and TorchScript. In our experiments, we observe slight advantages of TensorRT for convolutional layers and TorchScript for fully connected layers. We also study the trade-off between runtime and performance, when selecting an optimized setup for deployment, and observe that quantization significantly reduces the runtime while having only little impact on the detection performance.
3D point cloud understanding has made great progress in recent years. However, one major bottleneck is the scarcity of annotated real datasets, especially compared to 2D object detection tasks, since a large amount of labor is involved in annotating the real scans of a scene. A promising solution to this problem is to make better use of the synthetic dataset, which consists of CAD object models, to boost the learning on real datasets. This can be achieved by the pre-training and fine-tuning procedure. However, recent work on 3D pre-training exhibits failure when transfer features learned on synthetic objects to other real-world applications. In this work, we put forward a new method called RandomRooms to accomplish this objective. In particular, we propose to generate random layouts of a scene by making use of the objects in the synthetic CAD dataset and learn the 3D scene representation by applying object-level contrastive learning on two random scenes generated from the same set of synthetic objects. The model pre-trained in this way can serve as a better initialization when later fine-tuning on the 3D object detection task. Empirically, we show consistent improvement in downstream 3D detection tasks on several base models, especially when less training data are used, which strongly demonstrates the effectiveness and generalization of our method. Benefiting from the rich semantic knowledge and diverse objects from synthetic data, our method establishes the new state-of-the-art on widely-used 3D detection benchmarks ScanNetV2 and SUN RGB-D. We expect our attempt to provide a new perspective for bridging object and scene-level 3D understanding.
One-stage object detection is commonly implemented by optimizing two sub-tasks: object classification and localization, using heads with two parallel branches, which might lead to a certain level of spatial misalignment in predictions between the two tasks. In this work, we propose a Task-aligned One-stage Object Detection (TOOD) that explicitly aligns the two tasks in a learning-based manner. First, we design a novel Task-aligned Head (T-Head) which offers a better balance between learning task-interactive and task-specific features, as well as a greater flexibility to learn the alignment via a task-aligned predictor. Second, we propose Task Alignment Learning (TAL) to explicitly pull closer (or even unify) the optimal anchors for the two tasks during training via a designed sample assignment scheme and a task-aligned loss. Extensive experiments are conducted on MS-COCO, where TOOD achieves a 51.1 AP at single-model single-scale testing. This surpasses the recent one-stage detectors by a large margin, such as ATSS (47.7 AP), GFL (48.2 AP), and PAA (49.0 AP), with fewer parameters and FLOPs. Qualitative results also demonstrate the effectiveness of TOOD for better aligning the tasks of object classification and localization. Code is available at https://github.com/fcjian/TOOD.
In this report, we introduce our real-time 2D object detection system for the realistic autonomous driving scenario. Our detector is built on a newly designed YOLO model, called YOLOX. On the Argoverse-HD dataset, our system achieves 41.0 streaming AP, which surpassed second place by 7.8/6.1 on detection-only track/fully track, respectively. Moreover, equipped with TensorRT, our model achieves the 30FPS inference speed with a high-resolution input size (e.g., 1440-2304). Code and models will be available at https://github.com/Megvii-BaseDetection/YOLOX
Most of the existing single-stage and two-stage 3D object detectors are anchor-based methods, while the efficient but challenging anchor-free single-stage 3D object detection is not well investigated. Recent studies on 2D object detection show that the anchor-free methods also are of great potential. However, the unordered and sparse properties of point clouds prevent us from directly leveraging the advanced 2D methods on 3D point clouds. We overcome this by converting the voxel-based sparse 3D feature volumes into the sparse 2D feature maps. We propose an attentive module to fit the sparse feature maps to dense mostly on the object regions through the deformable convolution tower and the supervised mask-guided attention. By directly regressing the 3D bounding box from the enhanced and dense feature maps, we construct a novel single-stage 3D detector for point clouds in an anchor-free manner. We propose an IoU-based detection confidence re-calibration scheme to improve the correlation between the detection confidence score and the accuracy of the bounding box regression. Our code is publicly available at \url{https://github.com/jialeli1/MGAF-3DSSD}.