Direct Object Recognition Without Line-Of-Sight Using Optical Coherence

Lei, Xin; He, Liangyu; Tan, Yizheng; Wang, Ken Xingze; Wang, Xinggang; Du, Yihan; Fan, Shanhui; Yu, Zongfu

doi:10.1109/cvpr.2019.01201

Cited by 22 publications

(14 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Therefore, regardless of the light intensity, we assume that the full dynamic range of a 12-bit camera is used. To vary the noise, we keep σ 2 d fixed and change | F (u, v)| 2 to study the peak SNR (PSNR), 58 P SN R = 10 log 10 P pk,signal P noise (15) = 10 log 10…”

Section: Speed and Robustness To Noisementioning

confidence: 99%

“…"Non-line-of-sight" CNN imaging is recently demonstrated from albedo autocorrelation patterns of speckles. [13][14][15] In these examples, as well as others that leverage neural networks, it is possible to reconstruct an object type without solving the Phase Problem, i.e., one may successfully predict an object using prior-trained patterns without being able to identify the position of the object. 13 Still, deep learning neural networks offer additional functionality in the process of reconstructing the object.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Small-brain neural networks rapidly solve inverse problems with vortex Fourier encoders

Muminov¹,

Vuong²

2020

Preprint

View full text Add to dashboard Cite

We introduce a vortex phase transform with a lenslet-array to accompany shallow, dense, "small-brain" neural networks for high-speed and low-light imaging. Our single-shot ptychographic approach exploits the coherent diffraction, compact representation, and edge enhancement of Fourier-tranformed spiral-phase gradients. With vortex spatial encoding, a small brain is trained to deconvolve images at rates 5-20 times faster than those achieved with random encoding schemes, where greater advantages are gained in the presence of noise. Once trained, the small brain reconstructs an object from intensity-only data, solving an inverse mapping without performing iterations on each image and without deep-learning schemes. With this hybrid, optical-digital, vortex Fourier encoded, small-brain scheme, we reconstruct MNIST Fashion objects illuminated with low-light flux (5 nJ/cm 2 ) at a rate of several thousand frames per second on a 15 W central processing unit, two orders of magnitude faster than convolutional neural networks.

show abstract

Section: Speed and Robustness To Noisementioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Small-brain neural networks rapidly solve inverse problems with vortex Fourier encoders

Muminov¹,

Vuong²

2020

Preprint

View full text Add to dashboard Cite

show abstract

“…Our algorithm is deep learning based and consists of two neural networks: a compressor and a generator [30]. There are some pioneering works that have introduced deep learning into NLOS imaging and revealed the promising future of deep learning assisted NLOS imaging [13,14,32,34,2]. The reconstruction speeds of deep learning approaches are inherently much faster than geometric optics-based methods, with better versatility and generalization abilities.…”

Section: Introductionmentioning

confidence: 99%

Fast Non-line-of-sight Imaging with Two-step Deep Remapping

Zhu¹,

Cai²

2021

Preprint

View full text Add to dashboard Cite

Conventional imaging only records the photons directly sent from the object to the detector, while non-line-of-sight (NLOS) imaging takes the indirect light into account. To explore the NLOS surroundings, most NLOS solutions employ a transient scanning process, followed by a back-projection based algorithm to reconstruct the NLOS scenes. However, the transient detection requires sophisticated apparatus, with long scanning time and low robustness to ambient environment, and the reconstruction algorithms typically cost tens of minutes with high demand on memory and computational resources. Here we propose a new NLOS solution to address the above defects, with innovations on both detection equipment and reconstruction algorithm. We apply inexpensive commercial Lidar for detection, with much higher scanning speed and better compatibility to real-world imaging tasks. Our reconstruction framework is deep learning based, consisting of a variational autoencoder and a compression neural network. The generative feature and the two-step reconstruction strategy of the framework guarantee high fidelity of NLOS imaging. The overall detection and reconstruction process allows for real-time responses, with state-of-the-art reconstruction performance. We have experimentally tested the proposed solution on both a synthetic dataset and real objects, and further demonstrated our method to be applicable for full-color NLOS imaging.

show abstract

“…With the rapid development of deep learning in recent years, deep learning has been widely applied in image processing and provides a new method for speckle images analysis. This technology has been used to correctly recognize object located in the blind zone of the visual field [17], [18] and image reconstruction [19], [20] by analyzing the speckle patterns of objects. The customized neural network can realize the classification by learning the speckle image feature formed by irradiating the target image through the rough solid scattering medium [21] or turbid liquid [22].…”

Section: Introductionmentioning

confidence: 99%