Action Recognition From Single Timestamp Supervision in Untrimmed Videos
By Davide Moltisanti, Sanja Fidler, Dima Damen. 2019
69 citations recorded by the source corpus.
Recognising actions in videos relies on labelled supervision during training, typically the start and end times of each action instance. This supervision is not only subjective, but also expensive to acquire. Weak video-level supervision has been successfully exploited for recognition in untrimmed videos, however it is challenged when the number of different actions in training videos increases. We propose a method that is supervised by single timestamps located around each action instance, in untrimmed videos. We replace expensive action bounds with sampling distributions initialised from these timestamps. We then use the classifier's response to iteratively update the sampling distributions. We demonstrate that these distributions converge to the location and extent of discriminative action segments. We evaluate our method on three datasets for fine-grained recognition, with increasing number of different actions per video, and show that single timestamps offer a reasonable compromise between recognition performance and labelling effort, performing comparably to full temporal supervision. Our update method improves top-1 test accuracy by up to 5.4%. across the evaluated datasets.
Subfields: Computer Vision and Pattern Recognition
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computer Vision and Pattern Recognition
- Giovanni Maria Farinella
- Antonino Furnari
- Kristen Grauman
- Dima Damen
- Jitendra Malik
- Luc Van Gool
- Andrew Zisserman
- Cordelia Schmid
Citations in the corpus
- Rescaling Egocentric Vision: Collection, Pipeline and Challenges for EPIC-KITCHENS-100 (2020, background): “We use the definition from [63].”
- The EPIC-KITCHENS Dataset: Collection, Challenges and Baselines (2020, background): “This shows the challenge involved in annotating actions starting from narration timestamps, as well as the challenges involved in using these timestamps as weak temporal supervision as recently showed in [42].”