Iterated Second-Order Label Sensitive Pooling for 3D Human Pose Estimation

By Catalin Ionescu, João Carreira, Cristian Sminchisescu. 2014

104 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, 3D Vision & Pose, Human-Computer Interaction

How researchers see this paper

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

In Computer Vision and Pattern Recognition

In 3D Vision & Pose

In Human-Computer Interaction

Citations in the corpus

Hypercolumns for object segmentation and fine-grained localization (2014, neutral): “[25] jointly infer part segmentations and pose.”
Space-Time Representation of People Based on 3D Skeletal Data: A Review (2016, background): “72] Geodesic distance & optical ﬂow RGBD image streams 3D skeleton, 16 joints, robust to occlusions [73] energy cost optimization with multi-constraints Depth image sequence model per frame = 3ms [74] Second-order label-sensitive pooling RGB images 3D pose, 106mm precision [75] Recurrent 2D/3D p”
Human Pose Estimation with Iterative Error Feedback (2015, methodology): “The system in our human pose estimation example can be seen as closest to approaches employing “pose-indexed features” [12, 8, 17], but leveraging hierarchical feature learning.”
Learning Monocular 3D Human Pose Estimation from Multi-view Images (2018, methodology): “In the top portion of Table 1, we report the MPJPE and NMPJPE values of fully-supervised methods on H36M using the same protocol as in [13, 31, 16, 26, 18, 19]: Five subjects (S1, S5, S6, S7, S8) are used for training and the remaining two (S9, S11) for testing.”
Hands Deep in Deep Learning for Hand Pose Estimation (2015, background): “[9] recently showed it was possible to do the same from RGB images only, by combined body part labelling and iterative structured-output regression for 3D joint localization.”
Keep It SMPL: Automatic Estimation of 3D Human Pose and Shape from a Single Image (2016, methodology): “While not directly comparable, Ionescu et al. [17] report an error of 92mm on this dataset.”
Learning Camera Viewpoint Using CNN to Improve 3D Body Pose Estimation (2016, methodology): “For example, in [14] first body parts are segmented and then are described by second-order label-sensitive pooling [6]; the approach in [4] represents the image with HOG features; and LAB and HOG features are used in [16].”
Synthesizing Training Images for Boosting Human 3D Pose Estimation (2016, unknown): “[19] recover 3D pose with depth information.”
Fusing 2D Uncertainty and 3D Cues for Monocular Body Pose Estimation (2016, background): “Discriminative methods aim at predicting 3D pose directly from the input data, may it be single images [26, 27, 34, 35, 36, 46, 49, 58, 67], depth images [22, 44, 52], or short image sequences [59].”
Monocular 3D Human Pose Estimation in the Wild Using Improved CNN Supervision (2016, methodology): “3D annotation [23] is harder to obtain than 2D pose annotation.”
Learning from Synthetic Humans (2017, methodology): “D/3D positions of body joints. 3.2. Generating ground truth for real human data Human3.6M dataset [17,18] provides ground truth for 2D and 3D human poses. Additionally, a subset of the dataset (H80K) [16] has segmentation annotation, but the deﬁnition of parts is different from the SMPL body parts u”
Learning to Fuse 2D and 3D Image Cues for Monocular Body Pose Estimation (2016, unknown): “Discriminative methods aim at predicting 3D pose directly from the input data, may it be single images [28, 29, 37, 38, 39, 46, 52, 55, 64, 74], depth images [23, 50, 59], or short image sequences [65].”
Spatial Attention Deep Net with Partial PSO for Hierarchical Hybrid Hand Pose Estimation (2016, background): “Many prior works have achieved good performance by different methods [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16].”
Unite the People: Closing the Loop Between 3D and 2D Human Representations (2017, background): “4.2) from the fits to the original images, we obtain consistent labels while retaining generalization performance.”
MixSTE: Seq2seq Mixed Spatio-Temporal Encoder for 3D Human Pose Estimation in Video (2022, methodology): “[0] Hip [1] R Hip [2] R Knee [3] R Foot [4] L Hip [5] L Knee [6] L Foot [7] Spine [8] Thorax [9] Neck [10] Head [11] L Shoulder [12] L Elbow [13] L Wrist [14] R Shoulder [15] R Elbow [16] R Wrist”
VNect (2017, unknown): “Request permissions from permissions@acm.org.”
Beyond Bounding Boxes: Precise Localization of Objects in Images (2015, unknown): “[35] jointly infer part segmentations and pose.”
A Simple Yet Effective Baseline for 3d Human Pose Estimation (2017, background): “Another way of compiling knowledge about 3d human pose from datasets is by creating overcomplete bases suitable for representing human poses as sparse combinations [2, 7, 36, 49, 55, 56], lifting the pose to a reproducible kernel Hilbert space (RHKS) [18] or by creating novel priors from specialized”