Limits...
Forest Walk Methods for Localizing Body Joints from Single Depth Image.

Jung HY, Lee S, Heo YS, Yun ID - PLoS ONE (2015)

Bottom Line: A regression forest is trained to estimate the probability distribution to the direction or offset toward the particular joint, relative to the adjacent position.The distribution for next position is found from traversing the regression tree from new position.The continual position sampling through 3D space will eventually produce an expectation of sample positions, which we estimate as the joint position.

View Article: PubMed Central - PubMed

Affiliation: Division of Computer and Electronic Systems Engineering, Hankuk University of Foreign Studies, Yongin, Republic of Korea.

ABSTRACT
We present multiple random forest methods for human pose estimation from single depth images that can operate in very high frame rate. We introduce four algorithms: random forest walk, greedy forest walk, random forest jumps, and greedy forest jumps. The proposed approaches can accurately infer the 3D positions of body joints without additional information such as temporal prior. A regression forest is trained to estimate the probability distribution to the direction or offset toward the particular joint, relative to the adjacent position. During pose estimation, the new position is chosen from a set of representative directions or offsets. The distribution for next position is found from traversing the regression tree from new position. The continual position sampling through 3D space will eventually produce an expectation of sample positions, which we estimate as the joint position. The experiments show that the accuracy is higher than current state-of-the-art pose estimation methods with additional advantage in computation time.

No MeSH data available.


Related in: MedlinePlus

(a) mAP for EVAL set. (b) mAP of each joint for EVAL set. (c) mAP on SMMC-10 set. The proposed approach is compared with the recent algorithms using EVAL [3] set in (a) and (b). (a) shows our 3 results in red with different fps. RTW approach performs slightly higher than Ye and Yang [32] at 1262 fps. Even at 2687 fps, RTW shows higher precision than Ganapathi et al. [3]. The precisions in each joint are presented in (b). (c) compares results for SMMC-10 test sequence. Shotton et al. [5], Girshick et al. [10] and our approach are methods for pose estimation from a single image.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4581738&req=5

pone.0138328.g006: (a) mAP for EVAL set. (b) mAP of each joint for EVAL set. (c) mAP on SMMC-10 set. The proposed approach is compared with the recent algorithms using EVAL [3] set in (a) and (b). (a) shows our 3 results in red with different fps. RTW approach performs slightly higher than Ye and Yang [32] at 1262 fps. Even at 2687 fps, RTW shows higher precision than Ganapathi et al. [3]. The precisions in each joint are presented in (b). (c) compares results for SMMC-10 test sequence. Shotton et al. [5], Girshick et al. [10] and our approach are methods for pose estimation from a single image.

Mentions: The recent state-of-the-art pose tracking algorithm implementation by Ye and Yang [32] showed 0.921 mAP on EVAL sequences [32]. Their implementation achieved faster than 30 fps operation using GPU. The pose tracking method by Ganapathi et al. [3] does not rely on GPU to achieve 125 fps operation, but the precision is significantly lower. In Fig 6a, the proposed GFW is compared with these two methods [3, 32]. Our method is able to obtain slightly higher precision than Ye and Yang’s state-of-the-art tracking algorithm [32]. In addition, the computation is more than 40 times faster than their GPU assisted implementation [32].


Forest Walk Methods for Localizing Body Joints from Single Depth Image.

Jung HY, Lee S, Heo YS, Yun ID - PLoS ONE (2015)

(a) mAP for EVAL set. (b) mAP of each joint for EVAL set. (c) mAP on SMMC-10 set. The proposed approach is compared with the recent algorithms using EVAL [3] set in (a) and (b). (a) shows our 3 results in red with different fps. RTW approach performs slightly higher than Ye and Yang [32] at 1262 fps. Even at 2687 fps, RTW shows higher precision than Ganapathi et al. [3]. The precisions in each joint are presented in (b). (c) compares results for SMMC-10 test sequence. Shotton et al. [5], Girshick et al. [10] and our approach are methods for pose estimation from a single image.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4581738&req=5

pone.0138328.g006: (a) mAP for EVAL set. (b) mAP of each joint for EVAL set. (c) mAP on SMMC-10 set. The proposed approach is compared with the recent algorithms using EVAL [3] set in (a) and (b). (a) shows our 3 results in red with different fps. RTW approach performs slightly higher than Ye and Yang [32] at 1262 fps. Even at 2687 fps, RTW shows higher precision than Ganapathi et al. [3]. The precisions in each joint are presented in (b). (c) compares results for SMMC-10 test sequence. Shotton et al. [5], Girshick et al. [10] and our approach are methods for pose estimation from a single image.
Mentions: The recent state-of-the-art pose tracking algorithm implementation by Ye and Yang [32] showed 0.921 mAP on EVAL sequences [32]. Their implementation achieved faster than 30 fps operation using GPU. The pose tracking method by Ganapathi et al. [3] does not rely on GPU to achieve 125 fps operation, but the precision is significantly lower. In Fig 6a, the proposed GFW is compared with these two methods [3, 32]. Our method is able to obtain slightly higher precision than Ye and Yang’s state-of-the-art tracking algorithm [32]. In addition, the computation is more than 40 times faster than their GPU assisted implementation [32].

Bottom Line: A regression forest is trained to estimate the probability distribution to the direction or offset toward the particular joint, relative to the adjacent position.The distribution for next position is found from traversing the regression tree from new position.The continual position sampling through 3D space will eventually produce an expectation of sample positions, which we estimate as the joint position.

View Article: PubMed Central - PubMed

Affiliation: Division of Computer and Electronic Systems Engineering, Hankuk University of Foreign Studies, Yongin, Republic of Korea.

ABSTRACT
We present multiple random forest methods for human pose estimation from single depth images that can operate in very high frame rate. We introduce four algorithms: random forest walk, greedy forest walk, random forest jumps, and greedy forest jumps. The proposed approaches can accurately infer the 3D positions of body joints without additional information such as temporal prior. A regression forest is trained to estimate the probability distribution to the direction or offset toward the particular joint, relative to the adjacent position. During pose estimation, the new position is chosen from a set of representative directions or offsets. The distribution for next position is found from traversing the regression tree from new position. The continual position sampling through 3D space will eventually produce an expectation of sample positions, which we estimate as the joint position. The experiments show that the accuracy is higher than current state-of-the-art pose estimation methods with additional advantage in computation time.

No MeSH data available.


Related in: MedlinePlus