Acta Optica Sinica, Volume. 39, Issue 7, 0711002(2019)
Non-Line-of-Sight Imaging Through Deep Learning
Fig. 4. Structures of bottleneck layer. (a) Classical structure of bottleneck layer. (b) improved structure of bottleneck layer
Fig. 5. Retrieving results of different bottleneck layer models. (a) Ground truths; (b) speckle images; (c) retrieving results of classical bottleneck layer model; (d) retrieving results of improved bottleneck layer model; (i)-(iv) represent different handwriting characters 3,4,5, and 8, respectively
Fig. 6. Retrieving results of handwriting English letters. (a) Ground truths; (b) speckle images; (c) retrieving results of URNet model; (i)-(iv) represent different handwriting English letters U, S, H, and C,respectively
Fig. 7. Retrieving results under external interference. (a) Results of occlusion of speckle signal; (b) retrieving results of URNet model. (i)-(vi) represent occlusion proportion of 0, 0.0625, 0.1250, 0.1875, 0.2500, and 0.3125, respectively
Fig. 8. Retrieving results of different models. (a) Ground truths; (b) retrieving results of URNet model; (c) retrieving results of speckle autocorrelation model; (i)-(iv) different handwriting characters 9, 2, 6, and 0, respectively
Get Citation
Copy Citation Text
Tingyi Yu, Mu Qiao, Honglin Liu, Shensheng Han. Non-Line-of-Sight Imaging Through Deep Learning[J]. Acta Optica Sinica, 2019, 39(7): 0711002
Category: Imaging Systems
Received: Jan. 29, 2019
Accepted: Mar. 22, 2019
Published Online: Jul. 16, 2019
The Author Email: Liu Honglin (hlliu4@hotmail.com)