TY - GEN
T1 - Depthwise Over-Parameterized CNN for Voxel Human Pose Classification
AU - Putra, Oddy Virgantara
AU - Riandini,
AU - Yuniarno, Eko Mulyanto
AU - Purnomo, Mauridhi Hery
N1 - Publisher Copyright:
© 2023 IEEE.
PY - 2023
Y1 - 2023
N2 - Light Detection and Ranging (LiDAR) capture objects and backgrounds using a laser sensor, producing unstructured points in 3-dimensional called point clouds (PC). However, captured human pose PC is limited partially due to the LiDAR scan. The only information in the scanned area exists. Due to the inadequacy of PC data, it is challenging to classify such data. In this paper, we proposed a solution to overcome those problems. It is a novel depthwise over-parameterized (DOConv) embedded into a simple CNN. The raw PCs are converted into a 3D voxel in the input layer. In the convolutional (Conv) layer, the regular Conv is substituted with a-three layered DOConv. Lastly, to assess the performance of our model, we commence an evaluation with multiple classifier algorithms in ModelNet40 and our human pose dataset. Accuracy, loss, recall, precision, F1-scores, and Geometric mean are engaged as performance indicators. To sum up, our model outperformed all compared classifiers in accuracy for the primary dataset by 87.06 % and ModelNet40 by 68.68%.
AB - Light Detection and Ranging (LiDAR) capture objects and backgrounds using a laser sensor, producing unstructured points in 3-dimensional called point clouds (PC). However, captured human pose PC is limited partially due to the LiDAR scan. The only information in the scanned area exists. Due to the inadequacy of PC data, it is challenging to classify such data. In this paper, we proposed a solution to overcome those problems. It is a novel depthwise over-parameterized (DOConv) embedded into a simple CNN. The raw PCs are converted into a 3D voxel in the input layer. In the convolutional (Conv) layer, the regular Conv is substituted with a-three layered DOConv. Lastly, to assess the performance of our model, we commence an evaluation with multiple classifier algorithms in ModelNet40 and our human pose dataset. Accuracy, loss, recall, precision, F1-scores, and Geometric mean are engaged as performance indicators. To sum up, our model outperformed all compared classifiers in accuracy for the primary dataset by 87.06 % and ModelNet40 by 68.68%.
KW - Depthwise Convolution
KW - Human Pose
KW - Lightweight CNN
KW - Over-parameterization
KW - Point Cloud
UR - http://www.scopus.com/inward/record.url?scp=85171199440&partnerID=8YFLogxK
U2 - 10.1109/ISITIA59021.2023.10221054
DO - 10.1109/ISITIA59021.2023.10221054
M3 - Conference contribution
AN - SCOPUS:85171199440
T3 - 2023 International Seminar on Intelligent Technology and Its Applications: Leveraging Intelligent Systems to Achieve Sustainable Development Goals, ISITIA 2023 - Proceeding
SP - 54
EP - 59
BT - 2023 International Seminar on Intelligent Technology and Its Applications
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 24th International Seminar on Intelligent Technology and Its Applications, ISITIA 2023
Y2 - 26 July 2023 through 27 July 2023
ER -
