Research Article
Large-Scale Coarse-to-Fine Object Retrieval Ontology and Deep Local Multitask Learning
Table 3
Summary of the contribution of each method for each criterion.
| Methods | Attribute learning model based on deep features with SVM classifiers (Figure 1) | Attribute learning model based on adaptive attribute domain with independent deep neural networks (Figure 2) | Attribute learning model based on the end-to-end deep neural network as a shared block with adaptive loss functions (Figure 3) | Our proposed deep local multitask learning |
| Inner-group correlation | Not focused | Focused | Focused | Focused | Intergroup correlation | Not focused | Focused | Focused | Not focused | Imbalanced data solving | Not focused | Not focused | Focused | Focused | Transfer learning | Feedforward only | Applied in each individual network | Not focused but can be applied | Focused | Large-scale data adaption | Not focused | Limited | Focused | Focused | Ontology | Not focused | Not focused but can be applied | Not focused but can be applied | Focused |
|
|