A more realistic and general setting for MTL and an architecture-agnostic algorithm to learn MTL on partially annotated data.
A novel model-agnostic training framework for scene graph generation based on the concept of label informativeness.
A task-adaptation technique and systematic analysis for cross-domain few-shot classification.
A universal representation learning technique for cross-domain few-shot classification.
A training set synthesis technique for condensing a large dataset into a small set of informative synthetic samples.
We propose to improve the performance of the state-of-the-art image captioning models by incorporating two sources of prior knowledge.
New transfer-based techniques for detecting object-level anomalies.