Learning Image Representations Tied to Egomotion from Unlabeled Video

By Dinesh Jayaraman, Kristen Grauman. 2017

24 citations recorded by the source corpus.

Subfields: Biophysics, 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 Biophysics

1. Limin Wang
2. Tianhao Li
3. Ruohan Gao
4. In-So Kweon
5. Donggeun Yoo
6. Dahun Kim
7. Zhirong Wu
8. Dahua Lin

In Computer Vision and Pattern Recognition

1. Luc Van Gool
2. Kristen Grauman
3. Dacheng Tao
4. Trevor Darrell
5. Jitendra Malik
6. Andrea Vedaldi
7. Andrew Zisserman
8. Ming-Hsuan Yang

Citations in the corpus

• Self-Supervised Learning of Video-Induced Visual Invariances (2019, neutral): “In a robotics setting, RGB pixels may be considered together with ego-motion [2, 30].”
• Unsupervised Feature Learning via Non-parametric Instance Discrimination (2018, methodology): “For videos, self-supervision strategies include: leveraging temporal continuity via tracking [44, 45], predicting future [42], or preserving the equivariance of egomotion [13, 50, 30].”
• Cross Pixel Optical Flow Similarity for Self-Supervised Learning (2018, neutral): “ by prediction audio from video frames [16,17]. [18] train a two stream architecture to classify whether an image and sound clip go together or not. Temporal information is coupled with ego-motion in [19,20]. [21] use videos along with spatial context pretraining [22] to construct an image graph. Tr”
• Learning Spatiotemporal Features via Video and Text Pair Discrimination (2020, methodology): “In video domain, typical examples include frame prediction [7, 54], optical flow estimation [37, 69, 23], instance tracking [59, 60], temporal order or structure prediction [36, 13, 61, 64].”
• ShapeCodes: Self-supervised Feature Learning by Lifting Views to Viewgrids (2017, methodology): “Furthermore, the ShapeCode provides a representation that is equivariant to egomotion transformations, which is known to benefit recognition and emerges naturally in supervised networks [44,57,58,59].”
• Learning Image Representations by Completing Damaged Jigsaw Puzzles (2018, methodology): “Recent methods commonly use images [4, 6, 21, 25, 26, 30, 38, 39], and often video [12, 24, 29, 36], or other sensory data such as egomotion and sound [1, 2, 13, 28].”
• Toward Driving Scene Understanding: A Dataset for Learning Driver Behavior and Causal Reasoning (2018, background): “Recently, learning a vision-based driving model [3, 12, 28, 34, 13] for autonomous driving has attracted a lot of attention.”