Since Monday 28th August, I had the priviledge to participate to the French national conference entitled GRETSI (Groupe de Recherche et d'Etudes de Traitement du Signal et des Images). The conference is particularly popular in the French signal and image processing community.

I had the opportunity to introduce my recent works about :
The Impact of the Decoding Strategy on Radar-to-Optical Modality Translation in Remote Sensing Imagery


The French version of the paper can be found following the link: https://gretsi.fr/data/colloque/pdf/2023_bralet1309.pdf

The study propose to evaluate the performances of SARDINet [Bralet et al, 2022] depending on the decoding strategy used. Five methods are compared:
  • Post-upsampling convolutions
  • Transposed convolutions
  • Sub-pixel convolutions from [Shi et al, 2016]
  • Post-upsampling convolutions with a final sub-pixel convolution layer
  • Transposed convolutions with a final sub-pixel convolution layer
I hope this work can be useful for you ! If so, please consider citing our work as follows :

Bralet, A., Atto, A., Chanussot, J., and Trouve, E. (2023). Impact de la stratégie de décodage sur la traduction de modalité radar-optique d’images de télédétection. In 29° Colloque sur le traitement du signal et des images, number 2023-1309, pages p. 929–932, Grenoble. GRETSI - Groupe de Recherche en Traitement du Signal et des Images

References

[Bralet et al, 2022] BRALET, Antoine, ATTO, Abdourrahmane M., CHANUSSOT, Jocelyn, et al. Deep Learning of Radiometrical and Geometrical Sar Distorsions for Image Modality translations. In : 2022 IEEE International Conference on Image Processing (ICIP). IEEE, 2022. p. 1766-1770.

[Shi et al, 2016] W. Shi et al., "Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network," 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, 2016, pp. 1874-1883, doi: 10.1109/CVPR.2016.207.