Determined by User Needs: A Salient Object Detection Rationale Beyond Conventional Visual Stimuli

Existing $\mathbf{\text{s}}$alient $\mathbf{\text{o}}$bject $\mathbf{\text{d}}$etection (SOD) methods adopt a $\mathbf{\text{passive}}$ visual stimulus-based rationale–objects with the strongest visual stimuli are perceived as the user’s primary focus (i.e., salient objects). They ignore the decisive role of users’ $\mathbf{\text{proactive}}\mathbf{\text{needs}}$ in segmenting salient objects–if a user has a need before seeing an image, the user’s salient objects align with their needs, e.g., if a user’s need is white apple'', when this user sees an image, the user's primary focus is on the white apple’’ or ``the most white apple-like’’ objects in the image. Such an oversight not only $\mathbf{\text{fails}}\mathbf{\text{to}}\mathbf{\text{satisfy}}\mathbf{\text{users}}$, but also $\mathbf{\text{limits}}\mathbf{\text{the}}\mathbf{\text{development}}\mathbf{\text{of}}\mathbf{\text{downstream}}\mathbf{\text{tasks}}$. For instance, in salient object ranking tasks, focusing solely on visual stimuli-based salient objects is insufficient for conducting an analysis of fine-grained relationships between users’ viewing order (usually determined by user’s needs) and scenes, which may result in wrong ranking results. Clearly, it is essential to detect salient objects based on user needs. Thus, we advocate a $\mathbf{\text{User}}$ $\mathbf{\text{S}}$alient $\mathbf{\text{O}}$bject $\mathbf{\text{D}}$etection (UserSOD) task, which focuses on $\mathbf{\text{detecting}}\mathbf{\text{salient}}\mathbf{\text{objects}}\mathbf{\text{align}}\mathbf{\text{with}}\mathbf{\text{users’}}\mathbf{\text{proactive}}\mathbf{\text{needs}}\mathbf{\text{when}}\mathbf{\text{user}}\mathbf{\text{have}}\mathbf{\text{needs}}$. The main challenge for this new task is the lack of datasets for model training and testing.