RGB-D object recognition and pose estimation based on pre-trained convolutional neural network features

By Max Schwarz, Hannes Schulz, Sven Behnke. 2015

346 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Control and Systems Engineering

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Sergey Levine
2. Li Fei-Fei
3. Silvio Savarese
4. Trevor Darrell
5. Chelsea Finn
6. Abhinav Kumar Gupta
7. Quoc V. Le
8. Ruslan Salakhutdinov

In Computer Vision and Pattern Recognition

1. Luc Van Gool
2. Jitendra Malik
3. Xiaogang Wang
4. Trevor Darrell
5. Dacheng Tao
6. Li Fei-Fei
7. Ming-Hsuan Yang
8. Andrea Vedaldi

In Control and Systems Engineering

1. Dieter Fox
2. Sergey Levine
3. Yuke Zhu
4. Li Fei-Fei
5. Silvio Savarese
6. Danfei Xu
7. Yu Xiang
8. Roberto Mart'in-Mart'in

Citations in the corpus

• CORe50: a New Dataset and Benchmark for Continuous Object Recognition (2017, neutral): “In the first group of datasets in Table 1 ( NORB , COIL-100 , iLAB20M , Washington RGB-D , BigBIRD , ALOI ), objects are positioned on turntables and acquisition is systematically controlled in term of pose/lighting.”
• 3D fully convolutional network for vehicle detection in point cloud (2016, methodology): “ucture of the FCN. B. Convolutional Neural Network and 3D Object Detection CNN based 3D object detection is recently drawing a growing attention in computer vision and robotics. [3], [9], [16], [17], [24], [25] embed 3D information in 2D projection and use 2D CNN for recognition or detection. [16] a”
• Vehicle Detection from 3D Lidar Using Fully Convolutional Network (2016, methodology): “For example on RGBD data, one common aspect is to treat the depthmaps as image channels and use 2D CNN for classification or detection [9, 24, 26].”
• PoseCNN: A Convolutional Neural Network for 6D Object Pose Estimation in Cluttered Scenes (2017, methodology): “With the emergence of depth cameras, several methods have been proposed for recognizing textureless objects using RGB-D data [13, 3, 2, 26, 15].”
• DenseFusion: 6D Object Pose Estimation by Iterative Dense Fusion (2019, methodology): “Other methods propose to directly estimate objects pose from images using CNN-based architectures [28, 36].”
• Online Continual Learning on Sequences (2020, background): “s. Frames per sess. Format Acquisition setting Outdoor sessions NORB [36] 5 25 20 20* grayscale turntable no COIL-100 [57] - 100 20 54* RGB turntable no iLab-20M [4] 15 704 - - RGB turntable no RGB-D [75] 51 300 - - RGB-D turntable no BigBIRD [78] - 100 - - RGB-D turntable no ALOI [16] - 1000 - - RG”
• Incremental On-Line Learning of Object Classes using a Combination of Self-Organizing Incremental Neural Networks and Deep Convolutional Neural Networks* (2016, methodology): “Some of these approaches have been tailored for the object recognition task by using a combination of RGB and depth information ([4], [5]) and fine-tuning or otherwise using pre-trained CNNs ([6], [7], [8]).”
• Joint Human Detection and Head Pose Estimation via Multistream Networks for RGB-D Videos (2017, background): “input [40] or extract discriminative features independently and then concatenate in the last fully connected layer [41], [42].”
• Incremental robot learning of new objects with fixed update time (2016, background): “1 on RGB-D is approximately 84%, which is comparable with the state of the art on this dataset [3].”
• Multimodal deep learning for robust RGB-D object recognition (2015, methodology): “[20] proposed to use features extracted from CNNs pre-trained on ImageNet for RGB-D object recognition.”
• iNeRF: Inverting Neural Radiance Fields for Pose Estimation (2020, methodology): “Recent approaches based on deep learning have proposed to 1) directly estimate objects pose using CNNbased architectures [32], [40], [46] or 2) estimate 2D keypoints [27], [35], [37], [38] and solve for pose using the PnP-RANSAC algorithm.”
• Robotic grasp detection using deep convolutional neural networks (2016, neutral): “[32].”
• Video Object Segmentation using Teacher-Student Adaptation in a Human Robot Interaction (HRI) Setting (2018, neutral): “RGB-D [25] - 51 300 T BIG BIRD [26] - - 100 T ICUB 28 [20] 4 7 28 H ICUB World [21] 6 20 200 H Core50 [17] 11 10 50 H IVOS 12 12 36 H”
• Incremental online learning of objects for robots operating in real environments (2017, methodology): “[9], the features obtained by the CNN are located in lower-dimensional pose manifolds where similar poses of an object remain close in the feature space as well.”
• Object identification from few examples by improving the invariance of a Deep Convolutional Neural Network (2016, background): “standard classifier such as a Support Vector Machine (SVM) or a Regularized Least Squares Classifier (RLS) [16] on such features consistently outperforms more sophisticated competitors on a variety of challenging datasets [17], [3], [18], [19], [20], [10], [21].”
• Knowledge is Never Enough: Towards Web Aided Deep Open World Recognition (2019, background): “abilities of intelligent autonomous systems in the last years [7], [8], [9], [10].”
• Progress in Neuroengineering for brain repair: New challenges and open issues (2018, background): “A similar problem is being studied in robotics for training visual recognition systems from small dataset (Pasquale et al., 2016; Schwarz et al., 2015).”