Leveraging Uncertainty to Rethink Loss Functions and Evaluation Measures for Egocentric Action Anticipation
By Antonino Furnari, Sebastiano Battiato, Giovanni Maria Farinella. 2018
77 citations recorded by the source corpus.
Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Probabilistic & Bayesian Methods
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Giovanni Maria Farinella
- Antonino Furnari
- Arsha Nagrani
- Cordelia Schmid
- Paul Hongsuck Seo
- Dima Damen
- Angela Yao
- Davide Moltisanti
In Computer Vision and Pattern Recognition
- Giovanni Maria Farinella
- Antonino Furnari
- Kristen Grauman
- Cordelia Schmid
- Luc Van Gool
- Jitendra Malik
- Andrew Zisserman
- Dima Damen
In Probabilistic & Bayesian Methods
- Giovanni Maria Farinella
- Antonino Furnari
- Angela Yao
- Paul Hongsuck Seo
- Arsha Nagrani
- Cordelia Schmid
- Fadime Sener
Citations in the corpus
- Rescaling Egocentric Vision: Collection, Pipeline and Challenges for EPIC-KITCHENS-100 (2020, methodology): “EvaluationMetricsWe report results using class-mean top5 recall (Furnari et al. 2018).”
- Look Before you Speak: Visually Contextualized Utterances (2020, background): “There is also inherent ambiguity in the future – given the observation of the present, multiple predictions about the future are possible [23].”
- Rolling-Unrolling LSTMs for Action Anticipation from First-Person Video (2020, methodology): “Past approaches [9], [10], [11] performed action anticipation considering a fixed anticipation time τ a , usually set to 1 second.”
- What Would You Expect? Anticipating Egocentric Actions With Rolling-Unrolling LSTMs and Modality Attention (2019, methodology): “Differently from previous works which considered a fixed anticipation time [8, 16, 58], the proposed method can anticipate an action at multiple anticipation times, e.”
- Predicting the Future from First Person (Egocentric) Vision: A Survey (2021, background): “…example, [8] evaluates an example as a true positive, if the correct class is among the top 3 predictions, [153] consider the predicted trajectory giving the smallest error among the top 10 trajectories, [37] evaluate action antic-ipation using top-5 accuracy and [110] introduced mean top-5 recall.”
- Assembly101: A Large-Scale Multi-View Video Dataset for Understanding Procedural Activities (2022, result): “Similar to previous works [5, 9], we report class-mean Top-5 recall as it accounts for uncertainty in future predictions.”
- Anticipative Video Transformer (2021, unknown): “Recall refers to class-mean recall@5 from [22].”