IDA-3D: Instance-Depth-Aware 3D Object Detection From Stereo Vision for Autonomous Driving

Peng, Wanli; Pan, Hao; Liu, He; Sun, Yi

doi:10.1109/cvpr42600.2020.01303

Cited by 60 publications

(29 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Compared to previous RGB-based methods that only predict 3D bounding box representation and cannot output detailed and complete object surface normals, our approach complements them with shape reconstruction capability without sacrificing object detection performance. Instance-level modeling in 3D object detection builds a feature representation for an interested object in order to estimate its 3D attributes [34,37,43,44,54]. FQ-Net [43] draws a re-projected 3D cuboid on an instance patch to predict its 3D Intersection over Union (IoU) with the ground truth.…”

Section: Related Workmentioning

confidence: 99%

“…The binocular human visual system effortlessly perceives various 3D attributes of surrounding objects including their locations, orientations and detailed surface geometry. However, existing visual 3D object detection (3DOD) approaches are often limited to approximate object shapes as 3D bounding boxes [38,40,54,57]. A complete and detailed shape description for the detected objects is desirable as a complement to these 3D bounding boxes for various machine vision applications.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Joint stereo 3D object detection and implicit surface reconstruction

Cheng

2021

Preprint

View full text Add to dashboard Cite

Figure 1. Given a stereo pair of RGB images, S-3D-RCNN detects 3D objects and predict their implicit shapes with one forward pass. (a) Alpha-blended image pair to show the disparities. (b) 3D object predictions shown as 3D bounding boxes. (c) Shape predictions for the detected objects. (d) Estimated surface normal of the nearby object. (e) The predicted implicit shape supports spatially-varying resolution.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Joint stereo 3D object detection and implicit surface reconstruction

Cheng

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Some traditional methods [26,39] use region-based stereo matching to estimate object's depth, but their performance is poor because they are not based on deep learning. [30] predicts the depth of the target through the method of instance depth perception which can detect the three-dimensional box end-to-end, but it neglects the local structure information. Our approach introduces a structure-aware-depth-estimation module that directly predicts the depth of the 3D bounding box's center, and then rectifies the results by box estimation and dense alignment, which together benefit the accuracy of depth estimation and thus yield better 3D detection performance.…”

Section: Related Workmentioning

confidence: 99%

“…We evaluate the proposed SIDE method with KITTI 3D dataset [10]. Specifically, our AP 3D of car category is better than the state-of-the-art methods IDA-3D [30] in all kinds of cases with IoU=0.7. Especially, in the moderate and hard case, our method performs better than IDA-3D with over 3% AP 3D , which means our method can better detect objects far away or with large occlusions.…”

Section: Introductionmentioning

confidence: 99%

SIDE: Center-based Stereo 3D Detector with Structure-aware Instance Depth Estimation

Peng¹,

Zhu²,

Wang³

et al. 2021

Preprint

View full text Add to dashboard Cite

3D detection plays an indispensable role in environment perception. Due to the high cost of commonly used LiDAR sensor, stereo vision based 3D detection, as an economical yet effective setting, attracts more attention recently. For these approaches based on 2D images, accurate depth information is the key to achieve 3D detection, and most existing methods resort to a preliminary stage for depth estimation. They mainly focus on the global depth and neglect the property of depth information in this specific task, namely, sparsity and locality, where exactly accurate depth is only needed for these 3D bounding boxes. Motivated by this finding, we propose a stereo-image based anchor-free 3D detection method, called structure-aware stereo 3D detector (termed as SIDE), where we explore the instancelevel depth information via constructing the cost volume from RoIs of each object. Due to the information sparsity of local cost volume, we further introduce match reweighting and structure-aware attention, to make the depth information more concentrated. Experiments conducted on the KITTI dataset show that our method achieves the state-ofthe-art performance compared to existing methods without depth map supervision.

show abstract

“…The understanding of 3D properties of objects in the real world is critical for visionbased autonomous driving and traffic surveillance systems [1][2][3][4][5]. Compared with a twodimensional (2D) object detection task, the 3D object detection task involves nine degrees of freedom, in which the length, width, height, and pose of the 3D bounding box need to be detected.…”

Section: Introductionmentioning

confidence: 99%

Monocular 3D Object Detection Based on Uncertainty Prediction of Keypoints

Zhao

Liu

2021

Machines

View full text Add to dashboard Cite

Three-dimensional (3D) object detection is an important task in the field of machine vision, in which the detection of 3D objects using monocular vision is even more challenging. We observe that most of the existing monocular methods focus on the design of the feature extraction framework or embedded geometric constraints, but ignore the possible errors in the intermediate process of the detection pipeline. These errors may be further amplified in the subsequent processes. After exploring the existing detection framework of keypoints, we find that the accuracy of keypoints prediction will seriously affect the solution of 3D object position. Therefore, we propose a novel keypoints uncertainty prediction network (KUP-Net) for monocular 3D object detection. In this work, we design an uncertainty prediction module to characterize the uncertainty that exists in keypoint prediction. Then, the uncertainty is used for joint optimization with object position. In addition, we adopt position-encoding to assist the uncertainty prediction, and use a timing coefficient to optimize the learning process. The experiments on our detector are conducted on the KITTI benchmark. For the two levels of easy and moderate, we achieve accuracy of 17.26 and 11.78 in AP3D, and achieve accuracy of 23.59 and 16.63 in APBEV, which are higher than the latest method KM3D.

show abstract

IDA-3D: Instance-Depth-Aware 3D Object Detection From Stereo Vision for Autonomous Driving

Cited by 60 publications

References 28 publications

Joint stereo 3D object detection and implicit surface reconstruction

Joint stereo 3D object detection and implicit surface reconstruction

SIDE: Center-based Stereo 3D Detector with Structure-aware Instance Depth Estimation

Monocular 3D Object Detection Based on Uncertainty Prediction of Keypoints

Contact Info

Product

Resources

About