Grouping-Based Low-Rank Trajectory Completion and 3D Reconstruction
Authors: Katerina Fragkiadaki, Marta Salas, Pablo Arbelaez, Jitendra Malik
NeurIPS 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We present competitive results of our method on the recently proposed NRSf M benchmark of [17], under a fixed set of parameters and while handling incomplete trajectories, in contrast to existing approaches. Further, we present extensive reconstruction results in videos from two popular video segmentation benchmarks, VSB100 [18] and Moseg [19], that contain videos from Hollywood movies and Youtube. |
| Researcher Affiliation | Academia | Katerina Fragkiadaki EECS, University of California, Berkeley, CA 94720 katef@berkeley.edu; Marta Salas Universidad de Zaragoza, Zaragoza, Spain msalasg@unizar.es; Pablo Arbel aez Universidad de los Andes, Bogot a, Colombia pa.arbelaez@uniandes.edu.co; Jitendra Malik EECS, University of California, Berkeley, CA 94720 malik@eecs.berkeley.edu |
| Pseudocode | No | The paper describes the methods and algorithms in text and mathematical equations, but it does not contain structured pseudocode or algorithm blocks (e.g., a figure or section labeled 'Algorithm' or 'Pseudocode'). |
| Open Source Code | Yes | Our code is available at: www.eecs.berkeley.edu/ katef/nrsfm. |
| Open Datasets | Yes | We present competitive results of our method on the recently proposed NRSf M benchmark of [17]... Further, we present extensive reconstruction results in videos from two popular video segmentation benchmarks, VSB100 [18] and Moseg [19]... |
| Dataset Splits | No | The paper does not specify exact split percentages or sample counts for training, validation, and test sets. It primarily uses existing benchmarks for evaluation without detailing how their data might have been partitioned for any internal model development or parameter tuning beyond what the benchmarks might inherently define for their tasks. |
| Hardware Specification | No | The paper does not provide specific details about the hardware (e.g., GPU/CPU models, memory) used to run the experiments. |
| Software Dependencies | No | The paper mentions methods like 'dense optical flow tracking', 'motion trajectory clustering', and 'spectral clustering of dense point trajectories', referencing papers (e.g., [14], [19]), but does not list specific software dependencies with version numbers (e.g., Python, libraries, frameworks). |
| Experiment Setup | Yes | Our method uses exactly the same parameters and K = 9 for all four sequences. ... We used K = 8 for all sequences. ... For all videos we use K {1 5}. and κr is the truncated rank of W used for the Euclidean upgrade step. When κr > 3, we use the Euclidean upgrade proposed in [5]. κr = 3 gives the most stable face reconstruction results. |