Research Communication | Open Access
Volume 2021 | Communication ID 328
Volume 2021 | Communication ID 328
| Image-level to Pixel-level Labeling: A weakly supervised model |
Youssef Ouassit, Soufiane Ardchir, Mohamed Azzouazi, Mohammed Yassine El Ghoumari
Received |
Accepted |
Published |
March 01, 2021 |
March 10, 2021 |
March 15, 2021 |
Abstract: Semantic segmentation is a computer vision task that consists of assigning an object class to a set of pixels in an image, it clusters parts of the image together which belong semantically to the same object class. Due to the existence of sufficient labeled data, Deep learning has achieved the state of art in fully automatic segmentation [1], however in the medical field labeling process is expensive and time-consuming. We propose a weakly supervised model [2] where we infer object segmentation by leveraging only a few segmentation masks and object class information, and by considering only ...