Unite the People: Closing the Loop Between 3D and 2D Human Representations
Citations
1,462 citations
Cites methods or result from "Unite the People: Closing the Loop ..."
...We report the segmentation accuracy and average F1 score over all parts including the background as done in [20]....
[...]
...[20] take curated results from SMPLify to train 91 keypoint detectors corresponding to traditional body joints and points on the surface....
[...]
...We follow [5, 20] and use a regressor to obtain the 14 joints of Human3....
[...]
...We also evaluate our approach on the auxiliary task of human body segmentation on the 1000 test images of LSP [17] labeled by [20]....
[...]
...Existing methods for recovering 3D human mesh today focus on a multi-stage approach [5, 20]....
[...]
987 citations
Cites background or methods from "Unite the People: Closing the Loop ..."
...We use the code provided by [23] with both DeeperCut pose estimation landmark detector [18] for 14-landmark results and with the 91landmark alternative proposed in [23]....
[...]
...A semi-automated method is used for the ‘Unite the People’ (UP) dataset of [23], where human annotators verified the results of fitting the SMPL 3D deformable model [28] to 2D images....
[...]
...Surface-level supervision was only recently introduced for synthetic images in [45], while in [23] a dataset of 8515 images is annotated with keypoints and semi-automated fits of 3D models to images....
[...]
...The works of [23, 45] can be used as surrogates, but as we show in Sec....
[...]
...However, model fitting often fails in the presence of occlusions, or extreme poses, and is never guaranteed to be entirely successful – for instance, even after rejecting a large fraction of the fitting results, the feet are still often misaligned in [23]....
[...]
907 citations
725 citations
687 citations
Cites background or methods from "Unite the People: Closing the Loop ..."
...[39] uses silhouettes along with keypoints for the fitting algorithm....
[...]
...Tremendous progress has been made on estimating 3D human pose and shape from a single image [11, 21, 25, 29, 36, 37, 39, 48, 51]....
[...]
...Due to the lack of in-the-wild 3D ground-truth labels, these methods use weak supervision signals obtained from a 2D keypoint re-projection loss [29, 60, 62], use body/part segmentation as an intermediate representation [48, 51], or employ a human in the loop [39]....
[...]
References
15,935 citations
"Unite the People: Closing the Loop ..." refers methods in this paper
...Finegrained part segmentation has been added to the public parts of the VOC dataset [12] by Chen et al....
[...]
5,904 citations
"Unite the People: Closing the Loop ..." refers methods in this paper
...rom an image in 0.378s. The pose-predicting CNN is the computational bottleneck. Because our findings are not specific to a CNN model, we believe that by using a speed-optimized CNN, such as SqueezeNet [18], and further optimizations of the direct predictor, the proposed method could reach realtime speed. 5. Closing the Loop With the improved results for 3D fitting, which helped to create the dataset of ...
[...]
5,670 citations
3,865 citations
"Unite the People: Closing the Loop ..." refers background or methods in this paper
...While 2D keypoint prediction has seen considerable progress in the last years and could be considered close to being solved [19, 32, 42], 3D pose estimation from single images remains a challenge [4, 36, 44]....
[...]
...We use a state-of-the-art DeeperCut CNN [19] for our pose-related experiments, but believe that using other models such as Convolutional Pose Machines [42] or Stacked Hourglass Networks [32] would lead to similar findings....
[...]
...Their representational power has led to increasingly robust algorithms for bounding box detection [10], keypoint detection [19, 32, 42] and body part segmentation [7, 15, 43]....
[...]
...We use a state-of-the-art DeeperCut CNN [19] for our pose-related experiments, but believe that using other models such as Convolutional Pose Machines [43] or Stacked Hourglass Networks [33] would lead to similar findings....
[...]
3,170 citations
"Unite the People: Closing the Loop ..." refers background in this paper
...Their representational power has led to increasingly robust algorithms for bounding box detection [10], keypoint detection [19, 32, 42] and body part segmentation [7, 15, 43]....
[...]