For people reading: you can look at Backlinks or the Graph View on the right to find the pages about the papers I’ve been reading.

TIME TO READ

Topics to explore:

  • Unsupervised optical flow (including traditional energy techniques to give us lots of cool criteria to minimize)
  • Probabilistic formulations of optical flow
  • Supervised optical flow
  • Particle tracking
  • Diffusion modeling

Strategies:

  • Reading in chronological order: older papers are (usually) easier to understand (especially if you ignore the math)
  • Rereading a paper once I’ve read the ones it cited
  • I can look at Connected Papers to see some papers related to ones I’m currently working on

Origin papers:

Original LitReview doc: edit (with some supervised flow stuff I can paste over)

Non-local Total Generalized Variation for Optical Flow Estimation Pyramid methods in image processing learning-by-analogy-reliable-supervision-from unsupervised-optical-flow-using-cost-function 0a87428c6b2205240485ee6bb9cfb00fd9ed359c UnFlow Census Loss SelFLow Liu Lyu Xu 2cb35277b551f2ccc91f37aa66c2a8ae01be0c65

Secrets of optical flow estimation and their principles

GMFlow

Computation of optical flow using a neural network

[PDF] arxiv.org

Back to basics: Unsupervised learning of optical flow via brightness constancy and motion smoothness (2016!)

0b197f323e7d5514e5ada388858e38890dce0148 c90a2d5f881c70587643019e8af3c1fd0f6faa4d [

What Makes Good Synthetic Training Data for Learning Disparity and Optical _Flow_Estimation?

](ec1f3f1f1d9af3046650c3a30f95a7f2f0a78390) f5719438eb7e26827db9c3cd469063d1d12e1f19 Competitive Collaboration af3422d4f953ef54cd367cb48f7a28fba8e0addd 24466bb59d19206ce027729f238b68a848d47f94 15422a62664c6d79c9a3e70625e1f07b1f245c10 281c84088a4294bc0c74b4bb21d1800f2f925a5e GenerativeImageDynamics.pdf

Get true gt warped images oops

Find the best paper that currently exists and run it