Skip to Main content Skip to Navigation
Preprints, Working Papers, ...

Pointing is sufficient for remote sensing images

Abstract : There are a lot of works aiming to reduce the need of human annotations: self supervised training, interactive verification instead of annotation or weakly supervised training. For example, only pointing object centres is a weaker ground truth than providing bounding boxes or segmentation mask. Indeed, we show on several remote sensing public datasets that pointing can be sufficient for learning a segment before pointing pipeline, and, resulting pipeline also reaches honourable score on segmentation task.
Document type :
Preprints, Working Papers, ...
Complete list of metadata

Cited literature [22 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-01412086
Contributor : Adrien Chan-Hon-Tong Connect in order to contact the contributor
Submitted on : Friday, December 15, 2017 - 3:25:10 PM
Last modification on : Wednesday, November 3, 2021 - 7:36:43 AM

File

raw.pdf
Files produced by the author(s)

Identifiers

Citation

Adrien Chan-Hon-Tong, Nicolas Audebert. Pointing is sufficient for remote sensing images. 2017. ⟨hal-01412086v6⟩

Share

Metrics

Record views

260

Files downloads

360