**8. Conclusions**

We propose a novel CNN architecture to achieve image labeling on remote-sensed images. Our best-proposed method, "HR-GCN-FF-DA", delivers an excellent performance in regards to three aspects: (i) modifying the backbone architecture with "High-Resolution Representations (HR)", (ii) applying the "Feature Fusion (FF)", and (iii) using the concept of "Depthwise Atrous Convolution (DA)". Each proposed strategy can really improve *F*1-results by 4.82%, 4.08%, and 2.14% by adding HR, FF, and DA modules, consecutively. The FF module can really capture low-level features, resulting in a higher accuracy of river and low-vegetation classes. The DA module can refine the features and provide more coverage areas, resulting in a higher accuracy of pineapple and miscellaneous classes. The results demonstrate that the "HR-GCN-FF-DA" model significantly exceeds all baselines. It is the victor in all data sets and exceeds more than 90% of *F*1: 0.9114, 0.9362, and 0.9111 of the Landsat-8w3c, Landsat-8w5c, and ISPRS Vaihingen corpora, respectively. Moreover, it reaches an accuracy surpassing 90% in almost all classes.

**Author Contributions:** Conceptualization, T.P.; Data curation, K.J., S.L. and P.S.; Formal analysis, T.P.; Investigation, T.P.; Methodology, T.P.; Project administration, T.P.; Resources, T.P.; Software, T.P.; Supervision, T.P. and P.V.; Validation, T.P.; Visualization, T.P.; Writing–original draft, T.P.; Writing–review and editing, T.P. and P.V. All authors have read and agreed to the published version of the manuscript.

**Acknowledgments:** Teerapong Panboonyuen, also known as Kao Panboonyuen appreciates and thanks to the scholarship from the 100th Anniversary Chulalongkorn University Fund for the Doctoral Scholarship and the 90th Anniversary Chulalongkorn University Fund (Ratchadaphiseksomphot Endowment Fund. Teerapong Panboonyuen greatly acknowledges the Geo-Informatics and Space Technology Development Agency (GISTDA), Thailand, and Kao thanks to the staff from the GISTDA (Thanwarat Anan, Bussakon Satta, and Suwalak Nakya) for providing the remote sensing corpora used in this study.

**Conflicts of Interest:** The authors declare no conflict of interest.
