Actions and Attributes from Wholes and Parts
By Georgia Gkioxari, Ross B. Girshick, Jitendra Malik. 2014
151 citations recorded by the source corpus.
We investigate the importance of parts for the tasks of action and attribute classification. We develop a part-based approach by leveraging convolutional network features inspired by recent advances in computer vision. Our part detectors are a deep version of poselets and capture parts of the human body under a distinct set of poses. For the tasks of action and attribute classification, we train holistic convolutional neural networks and show that adding parts leads to top-performing results for both tasks. We observe that for deeper networks parts are less significant. In addition, we demonstrate the effectiveness of our approach when we replace an oracle person detector, as is the default in the current evaluation protocol for both tasks, with a state-of-the-art person detection system.
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
- Dacheng Tao
- Abhinav Kumar Gupta
- Jitendra Malik
- Trevor Darrell
- X. Wang
- Ross B. Girshick
- Joost van de Weijer
- Bernt Schiele
In Computer Vision and Pattern Recognition
- Dacheng Tao
- Luc Van Gool
- Ming-Hsuan Yang
- Xiaogang Wang
- Trevor Darrell
- Jitendra Malik
- Bernt Schiele
- Yi Yang
Citations in the corpus
- Recent advances in convolutional neural networks (2015, background): “[176] add a part detection to this framework.”
- Contextual Action Recognition with R*CNN (2015, methodology): “[12] and R∗CNN perform equally well, with the upside that R∗CNN does not need use keypoint annotations during training.”
- Do Semantic Parts Emerge in Convolutional Neural Networks? (2016, background): “Finally, note how future works that use filters as intermediate part representations (as in Simon et al. 2014; Gkioxari et al. 2015; Simon and Rodner 2015; Xiao et al. 2015; Oquab et al. 2015) will now be able to exploit these findings to create better models for recognition.”
- Videos as Space-Time Region Graphs (2018, background): “In the case of action recognition, a lot of effort has been made on modeling pairwise human-object and object-object relationships [55,56,57].”
- Part-Stacked CNN for Fine-Grained Visual Categorization (2015, background): “Recently, the best performing methods mostly sought for improvement brought by the following three aspects: more discriminative features including deep CNNs for better visual representation [5, 32, 19, 36, 34], explicit alignment approaches to eliminate pose displacements [6, 14], and part-based met”
- Semantic Image Interpretation - Integration of Numerical Data and Logical Knowledge for Cognitive Vision (2018, unknown): “They leverage the part-whole relation between whole objects and their parts to improve the object detection [17, 32, 35, 38, 32, 70].”
- SPDA-CNN: Unifying Semantic Part Detection and Abstraction for Fine-Grained Recognition (2016, methodology): “Currently, methods based on CNNs [41, 25, 28, 5, 15, 18] significantly outperform previous works that rely on handcrafted features for part detection, part abstraction and finegrained classification.”
- Recognizing Actions Through Action-Specific Person Detection (2015, methodology): “Motivated by the recent success of convolutional neural networks [16], [26], [31], we use deep features to represent our person proposals for action classification.”