Open Access
Table 3
A summary of the three usage methods
| Method | Strengths | Weaknesses |
|---|---|---|
| Multimodal feature fusion | Integrating the features of scene graph with observations can equip agent with cognition and spatial intelligence | The challenge of balancing the performance of network models with computational complexity |
| Value function estimation | Incorporation of scene graph into the decision-making layer, providing valuable insights into interject relationships for decision making | Overreliance on scene graph structure; Irrelevant information interference |
| Sparse reward improvement | Mitigating reward sparsity leveraging the spatial relations extracted from scene graph in conjunction with contrastive learning | Limited by task-specific requirements; The need for extensive negative samples due to contrastive learning |
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.