Bat-G net: Bat-inspired High-Resolution 3D Image Reconstruction using Ultrasonic Echoes
Gunpil Hwang, Seohyeon Kim, and Hyeon-Min Bae
Korea Advanced Institute of Science and Technology (KAIST)
Daejeon, South Korea
Abstract
In this paper, a bat-inspired high-resolution ultrasound 3D imaging system is presented. Live bats demonstrate that the properly used ultrasound can be used to perceive 3D space. With this in mind, a neural network referred to as a Bat-G network is implemented to reconstruct the 3D representation of target objects from the hyperbolic FM (HFM) chirped ultrasonic echoes. The Bat-G network consists of an encoder emulating a bat's central auditory pathway, and a 3D graphical visualization decoder. For the acquisition of the ultrasound data, a custom-made Bat-I sensor module is used. The Bat-G network shows the uniform 3D reconstruction results and achieves precision, recall, and F1-score of 0.896, 0.899 and 0.895, respectively. The experimental results demonstrate the implementation feasibility of a high-resolution non-optical sound-based imaging system being used by live bats.
Paper, Supplemental, Poster
Slides
DATASET
T000001_01_01_FU
T000001_01_01_FD
T000001_01_01_FL
T000001_01_01_FR
T000001_01_01_Label_depthmap
T000001_01_01_TU
T000001_01_01_TD
T000001_01_01_TL
T000001_01_01_TR
Measurement setup
Bat-inspired imaging (Bat-I) sensor emits broadband FM signals and records echoes reflected from the target object. The recorded data transformed into spectrogram are fed into the Bat-G network for training and the network eventually infers the object’s 3D representation. In order to train the network, we have adopted a supervised learning algorithm and created 4-channel ultrasound echo dataset, ECHO-4CH (49 k data for training and 2.6 k data for evaluation). Each echo data consists of eight spectrograms (256^2 grayscale image) and one 3D ground-truth label (64^3 voxels). The detailed description of ECHO-4CH dataset is covered in the Bat-G net paper.
- Measurement list and File numbering:
ex) T000001_05_02_FU
T000001 : Measurement list filenum, (CUBE, Position 1)
05 : Clockwise angle 50deg
02 : Repeat count 2
FU : High frequency resolution spectrogram from Up-side microphone (T: high time resolution spectrogram, F: high frequency resolution spectrogram / U: up, D: down, L: left, R: right)
--> Unlisted filenumbers were excluded due to low SNR
- Citation:
If you use this ECHO-4CH dataset, please add a reference using this bibtex entry.
@inproceedings{hwang2019bat,
title={Bat-G net: Bat-inspired High-Resolution 3D Image Reconstruction using Ultrasonic Echoes},
author={Hwang, Gunpil and Kim, Seohyeon and Bae, Hyeon-Min},
booktitle={Advances in Neural Information Processing Systems},
pages={3720--3731},
year={2019}
}