SP-ILC: Concurrent Single-Pixel Imaging, Object Location, and Classification by Deep Learning

Yang, Zhihan; Bai, Y.; Sun, Lida; Huang, Kexin; Li, Jun; Ruan, Dong; Li, Junlin

doi:10.3390/photonics8090400

Cited by 11 publications

(6 citation statements)

References 54 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It can be seen that the SA Net has good performance for scale estimation; for example, when the ŝ is 4, δ can reach 13.9% on the MNIST Large-Scale dataset. Besides, the SA Net also works on the FMNIST Large-Scale dataset, which are grayscale images and closer to real-world images compared to the MNIST [33]. Similar to the classification experiment, in order to select the optimal wavelet families, we comprehensively compare the scale estimation experimental results of the SA Net using Haar (haar), Daubechies (db) and Biorthogonal (bior).…”

Section: Experiments Resultsmentioning

confidence: 99%

Scale-Aware Network with Scale Equivariance

et al. 2022

View full text Add to dashboard Cite

The convolutional neural network (CNN) has achieved good performance in object classification due to its inherent translation equivariance, but its scale equivariance is poor. A Scale-Aware Network (SA Net) with scale equivariance is proposed to estimate the scale during classification. The SA Net only learns samples of one scale in the training stage; in the testing stage, the unknown-scale testing samples are up-sampled and down-sampled, and a group of image copies with different scales are generated to form the image pyramid. The up-sampling adopts interpolation, and the down-sampling adopts interpolation combined with wavelet transform to avoid spectrum aliasing. The generated test samples with different scales are sent to the Siamese network with weight sharing for inferencing. According to the position of the maximum value of the classification-score matrix, the testing samples can be classified and the scale can be estimated simultaneously. The results on the MNIST and FMNIST datasets show that the SA Net has better performance than the existing methods. When the scale is larger than 4, the SA Net has higher classification accuracy than other methods. In the scale-estimation experiment, the SA Net can achieve low relative RMSE on any scale. The SA Net has potential for effective use in remote sensing, optical image recognition and medical diagnosis in cytohistology.

show abstract

Section: Experiments Resultsmentioning

confidence: 99%

Scale-Aware Network with Scale Equivariance

et al. 2022

View full text Add to dashboard Cite

show abstract

“…Recently, image-free classification picks up interest again, due to the possibility of using neural network classifiers on the single-pixel signal [20,21]. Yang et al even proposed a scheme to classify, locate and reconstruct an image out of the single-pixel signal with one single neural network [22].…”

Section: Introductionmentioning

confidence: 99%

Evaluation of a time-gated-single-pixel-camera as a promising sensor for autonomous vehicles in harsh weather conditions

Bett¹

2023

J. Eur. Opt. Society-Rapid Publ.

View full text Add to dashboard Cite

We propose a Time-Gated-Single-Pixel-Camera as a promising sensor for image-free object detection for automotive application in adverse weather conditions. By combining the well-known principles of time-gating and single-pixel detection with neural networks, we aim to ultimately detect objects within the scene rapidly and robustly with a low-cost sensor. Here, we evaluate the possible data reduction such a system can provide compared to a conventional time-gated camera.

show abstract

“…Image-free object detection starts picking up interest with the development of deep neural networks 7,8 . Nevertheless, neural network architectures have one major draw-back for flexible object detection: The number of output nodes and thus the number of detectable objects must be fixed a priori.…”

Section: Introductionmentioning

confidence: 99%

“…Therefore, the data cannot be divided in grid cells. Indeed, to the best of our knowledge, all image-free solutions presented in literature so far circumvent the problem by using a priori information about the position and a sort of Yolo head -either by additionally reconstructing an image 7 or by scanning during measurement 8 . There exist first attemps to solve the problem of multiple objects iteratively 10 , yet in our case we prefer a single pass network due to the tight temporal constraint.…”

Section: Introductionmentioning

confidence: 99%

Towards image-free object detection for autonomous vehicles under harsh environmental conditions

Bett

Frenner

Reichelt

2023

Optical Measurement Systems for Industrial Inspection XIII

View full text Add to dashboard Cite

There exist two key aspects to enable autonomous vehicles: Robust sensors to provide all relevant data concerning the vehicle's surroundings as well as algorithms to evaluate this data in real time. Apart from radar and ultrasonic sensors, optical sensors such as lidar and cameras are state-of-the-art in prototype autonomous vehicles. In adverse weather conditions, such as e.g. fog, snow, dust, heavy rain or poorly illuminated scenes, though, those sensors do not perform reliably. Recently, we proposed to use a time-gated-single-pixel-camera to not only significantly reduce the amount of recorded data but additionally filter ballistic object photons, i.e. suppress the effect of noise from the obscuring medium. Apart from generating 3D object information, such a system can operate fast enough to deal with the highly dynamic environment as well as respect eye-safety norms. Moreover, a time-gated-single-pixel-camera offers the ability of image-free detection of all relevant objects within the scene which speeds up data evaluation as well. Here, we want to report on our progress towards realizing such a system. We will demonstrate image-free object detection on simulated data. We realize multi-object detection by generating object heat-maps for the different classes. Additionally, we discuss the difficulties we have to overcome to robustly detect objects in real measured data and shortly present our prototype setup, which we have implemented on a car together with our partners from Fraunhofer

show abstract

SP-ILC: Concurrent Single-Pixel Imaging, Object Location, and Classification by Deep Learning

Cited by 11 publications

References 54 publications

Scale-Aware Network with Scale Equivariance

Scale-Aware Network with Scale Equivariance

Evaluation of a time-gated-single-pixel-camera as a promising sensor for autonomous vehicles in harsh weather conditions

Towards image-free object detection for autonomous vehicles under harsh environmental conditions

Contact Info

Product

Resources

About