Watch-n-patch: Unsupervised understanding of actions and relations

By Chenxia Wu, Jiemi Zhang, Silvio Savarese, Ashutosh Saxena. 2015

135 citations recorded by the source corpus.

Subfields: Artificial Intelligence, 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 Artificial Intelligence

1. Juergen Gall
2. Wanqing Li
3. Pichao Wang
4. Philip O. Ogunbona
5. Alexander Richard
6. Hilde Kuehne
7. Juan Carlos Niebles
8. Li Fei-Fei

In Computer Vision and Pattern Recognition

1. Luc Van Gool
2. Wanqing Li
3. Li Fei-Fei
4. Juergen Gall
5. Pichao Wang
6. Jitendra Malik
7. Dacheng Tao
8. Cordelia Schmid

Citations in the corpus

• RGB-D-based action recognition datasets: A survey (2016, neutral): “RGB-D activity dataset The RGB-D activity dataset [37] (http://watchnpatch.”
• Space-Time Representation of People Based on 3D Skeletal Data: A Review (2016, neutral): “[180], [181], [183], [198], [209], [210].”
• PKU-MMD: A Large Scale Benchmark for Continuous Multi-Modal Human Action Understanding (2017, background): “Watch-n-Patch [51] and Compostable Activities [19] are the first datasets focusing on the continues sequences and the inner combination of activities in supervised or unsupervised methods.”
• Scene Flow to Action Map: A New Representation for RGB-D Based Action Recognition with Convolutional Neural Networks (2017, methodology): “These methods either extracted features from one modality: depth [49, 60, 29, 59, 24, 51, 52] or RGB [27, 33] or skeleton [43, 55, 56, 4, 34, 19], or fuse the features extracted separately from them at a later stage [28, 16, 10, 58].”
• Connectionist Temporal Modeling for Weakly Supervised Action Labeling (2016, methodology): “Our use of visual similarities in the training is related to unsupervised video parsing [28,29,45], where frames are grouped into segments based on visual or semantic cues.”
• A Review of Human Activity Recognition Methods (2015, methodology): “Then all scores are combined together in a supervised framework yielding a final decision score (Westerveld et al., 2003; Jiang et al., 2014).”
• Weakly supervised learning of actions from transcripts (2016, methodology): “It shows that our approach does well compared to the baseline and that it is able to detect actions in unseen sequences based on the weakly trained models.”
• Holistic++ Scene Understanding: Single-View 3D Holistic Scene Parsing and Human Pose Estimation With Human-Object Interaction and Physical Commonsense (2019, methodology): “For dataset without ground-truth 3D bounding boxes (i.e., Watch-n-Patch), we evaluate the distance between the camera center and the 3D object center.”
• Weakly Supervised Action Learning with RNN Based Fine-to-Coarse Modeling (2017, methodology): “Finally, [32] propose an unsupervised technique to derive action classes from RGB-D videos, respectively human skeleton representations, also considering an activity as a sequence of shortterm action clips.”
• BABEL: Bodies, Action and Behavior with English Labels (2021, unknown): “In [38], the authors introduce a test recorded by Kinect v2 in which they describe activities as compositions of action interactions with different objects.”
• LEMMA: A Multi-view Dataset for Learning Multi-agent Multi-task Activities (2020, background): “activity datasets on the basis of tasks and captured video contents; see a detailed summary in Table 1. Crowd-sourced from online videos and movie sharing platforms, typical largescale video datasets [53,29,8,9,15] focus on video-level summarization and classication . Although activity classes exhib”