Aniello Panariello
Aniello Panariello
Home
News
Posts
Publications
Projects
Contact
Light
Dark
Automatic
1
CLIP with Generative Latent Replay: a Strong Baseline for Incremental Learning
We propose a novel approach to mitigate forgetting while adapting a VLM, which exploits generative replay to align prompts to tasks. We also introduce a new metric to evaluate zero-shot capabilities within CL benchmarks.
Emanuele Frascaroli
,
Aniello Panariello
,
Pietro Buzzega
,
Lorenzo Bonicelli
,
Angelo Porrello
,
Simone Calderara
PDF
Cite
Code
Mask and Compress: Efficient Skeleton-based Action Recognition in Continual Learning
We introduce CHARON (Continual Human Action Recognition On skeletoNs), which maintains consistent performance while operating within an efficient framework. Through techniques like uniform sampling, interpolation, and a memory-efficient training stage based on masking, we achieve improved recognition accuracy while minimizing computational overhead.
Matteo Mosconi
,
Andriy Sorokin
,
Aniello Panariello
,
Angelo Porrello
,
Jacopo Bonato
,
Marco Cotogni
,
Luigi Sabetta
,
Simone Calderara
,
Rita Cucchiara
PDF
Cite
Code
TrackFlow: Multi-Object tracking with Normalizing Flows
We propose a novel approach to multi-object tracking that leverages Normalizing Flows to learn a joint probability distribution over the costs of candidate associations. Our experiments show that our approach consistently enhances the performance of several tracking-by-detection algorithms.
Gianluca Mancusi
,
Aniello Panariello
,
Angelo Porrello
,
Matteo Fabbri
,
Simone Calderara
,
Rita Cucchiara
PDF
Cite
Poster
Consistency-based Self-supervised Learning for Temporal Anomaly Localization
This work tackles Weakly Supervised Anomaly detection within the field of self-supervised learning, asking the model to yield the same anomaly scores for different augmentations of the same video sequence.
Aniello Panariello
,
Angelo Porrello
,
Simone Calderara
,
Rita Cucchiara
PDF
Cite
Code
Slides
DOI
Cite
×