PhD Dissertation Proposal by CHEN Zhaozheng | Weakly-Supervised Semantic Segmentation

Semantic segmentation is a fundamental task in computer vision that assigns a label to every pixel in an image based on the semantic meaning of the objects present. It demands a large amount of pixel-level labeled images for training deep models. Weakly-supervised semantic segmentation (WSSS) is a more feasible approach that uses only weak annotations to learn the segmentation task. Image-level label based WSSS is the most challenging and popular, where only the class label for the entire image is provided as supervision. To address this challenge, Class Activation Map (CAM) has emerged as a powerful technique in WSSS. It provides a way to visualize the areas of an image that are most relevant to a particular class without requiring pixel-level annotations. However, CAM is generated from the classification model and it often only highlights the most discriminative parts of the object due to the discriminative nature of the model. This dissertation examines the key issues behind conventional CAM and proposes corresponding solutions to address them. Our works focus on two crucial steps in CAM generation: training the classification model and computing CAM from the classification model.

CHEN Zhaozheng is a Ph.D. candidate at the School of Computing and Information Systems, Singapore Management University, supervised by Asst. Prof. SUN Qianru. His research focuses on weakly-supervised semantic segmentation.