Browsing by Author "Chen, Po-Heng"

Now showing 1 - 4 of 4

A Multi-band Frequency Synthesizer for GSM/DCS/WIMAX/WLAN Applications with Ripple-free Circuit
(2010)
Chen, Po-Heng
;
Chen, Che-Sheng
;
Chou, Mei-Fen
;
Wen, Kuei-Ann
;
電子工程學系及電子研究所
This paper presents a fully integrated multi- band frequency synthesizer implemented in CMOS 0.18 mu m process with ripple-free circuit in a single loop. The synthesizer is designed for GSM/ DCS/WIMAX/WLAN (802.11a/b/g) systems applying to IQ phase based transceivers. The proposed synthesizer with ripple-free circuit can reduce the amplitude of control voltage ripple. The synthesizer performs the suppression of 31.4dB in the spectrum of reference-spur and the peak-to-peak jitter with 74% improvement compared to the conventional single loop synthesizer design.
FADE: Feature Aggregation for Depth Estimation With Multi-View Stereo
(2020-01-01)
Yang, Hsiao-Chien
;
Chen, Po-Heng
;
Chen, Kuan-Wen
;
Lee, Chen-Yi
;
Chen, Yong-Sheng
Both structural and contextual information is essential and widely used in image analysis. However, current multi-view stereo (MVS) approaches usually use a single common pre-trained model as pixel descriptor to extract features, which mix structural and contextual information together and thus increase the difficulty of matching correspondence. In this paper, we propose FADE (feature aggregation for depth estimation), which treats spatial and context information separately and focuses on aggregating features for efficient learning of the MVS problem. Spatial information includes image details such as edges and corners, whereas context information comprises object features such as shapes and traits. To aggregate these multi-level features, we use an attention mechanism to select important features for matching. We then build a plane sweep volume by using a homography backward warping method to generate match candidates. Furthermore, we propose a novel cost volume regularization network aims to minimize the noise in the matching candidates. Finally, we take advantage of 3D stacked hourglass and regression to produces high-quality depth maps. With these well-aggregated features, FADE can efficiently perform dense depth reconstruction, achieving state-of-the-art performance in terms of accuracy and requiring the least amount of model parameters.
MVSNet plus plus : Learning Depth-Based Attention Pyramid Features for Multi-View Stereo
(2020-01-01)
Chen, Po-Heng
;
Yang, Hsiao-Chien
;
Chen, Kuan-Wen
;
Chen, Yong-Sheng
;
資訊工程學系
The goal of Multi-View Stereo (MVS) is to reconstruct 3D point-cloud model from multiple views. On the basis of the considerable progress of deep learning, an increasing amount of research has moved from traditional MVS methods to learning-based ones. However, two issues remain unsolved in the existing state-of-the-art methods: (1) only high-level information is considered for depth estimation. This may reduce the localization accuracy of 3D points as the learned model lacks spatial information; and (2) most of the methods require additional post-processing or network refinement to generate a smooth 3D model. This significantly increases the number of model parameters or the computational complexity. To this end, we propose MVSNet++, an end-to-end trainable network for dense depth estimation. Such an estimated depth map can further be applied to 3D model reconstruction. Different from previous methods, in the proposed method, we first adopt feature pyramid structures for both feature extraction and cost volume regularization. This can lead to accurate 3D point localization by fusing multi-level information. To generate smooth depth map, we then carefully integrate instance normalization into MVSNet++ without increasing model parameters and computational burden. Furthermore, we additionally design three loss functions and integrate Curriculum Learning framework into the training process, which can lead to an accurate reconstruction of 3D model. MVSNet++ is evaluated on DTU and Tanks & Temples benchmarks with comprehensive ablation studies. Experimental results demonstrate that our proposed method performs favorably against previous state-of-the-art methods, showing the accuracy and effectiveness of the proposed MVSNet++.
UAVNet: An Efficient Obstacel Detection Model for UAV with Autonomous Flight
(2018-01-01)
Chen, Po-Heng
;
Lee, Chen-Yi
;
電子工程學系及電子研究所
;
Department of Electronics Engineering and Institute of Electronics
Autonomous navigation for large Unmanned Aerial Vehicles(UAVs) is straight-forward to implement, just employ expensive and sophisticated sensors and monitoring devices. On the contrary, usual small quadrotor UAV still have the challenge on obstacle avoidance since this kind of UAV can only carry very light weight sensors such as cameras. Given the above reason, making autonomous navigation over obstacles on small UAV is much more challenging. In this paper, we focus on proposing a novel and memory efficient deep network architecture named UAVNet for small UAV to achieve obstacle detection in the urban environment. Compared with state-of-the-art DNN architecture, UAVNet has only 2.23M parameters(which is half compared with MobileNet) and 141 MFLOPs complexity. Though the parameters are fewer than usual, the accuracy is acceptable, about 80% validated on ImageNet-2102 dataset. To further justify the utility of UAVNet, we also implement the architecture on Nvidia TX2 in real environment using NCTU campus dataset. The experiment shows the proposed UAVNet can detect obstacles to 15 fps, which is a real-time application.