Encoding High Dimensional Local Features by Sparse Coding Based Fisher Vectors

Authors: Lingqiao Liu, Chunhua Shen, Lei Wang, Anton van den Hengel, Chao Wang

NeurIPS 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experimental evaluations demonstrate that our method not only significantly outperforms the traditional GMM based Fisher vector encoding but also achieves the state-of-the-art performance in generic object recognition, indoor scene, and fine-grained image classification problems.
Researcher Affiliation Academia 1 School of Computer Science, University of Adelaide, Australia 2 ARC Centre of Excellence for Robotic Vision 3 School of Computer Science and Software Engineering, University of Wollongong, Australia
Pseudocode No The paper includes mathematical derivations and formulations but does not present any structured pseudocode or algorithm blocks.
Open Source Code No The paper mentions using existing tools like “the Caffe [17] package” and “the algorithm in [18]”, but it does not provide any statement or link for the release of the authors’ own source code for the described methodology.
Open Datasets Yes We conduct experimental evaluation of the proposed sparse coding based Fisher vector coding (SCFVC) on three large datasets: Pascal VOC 2007, MIT indoor scene-67 and Caltech-UCSD Birds-200-2011. These are commonly used evaluation benchmarks for generic object classification, scene classification and fine-grained image classification respectively.
Dataset Splits No The paper uses standard datasets like Pascal VOC 2007, MIT-67, and Birds-200-2011, but it does not explicitly provide specific details on training, validation, and test dataset splits, such as percentages, sample counts, or explicit citations for predefined split methodologies.
Hardware Specification No The paper does not provide any specific details about the hardware (e.g., GPU models, CPU types, memory) used for running the experiments.
Software Dependencies No The paper mentions using “the Caffe [17] package” but does not provide a specific version number for this software dependency.
Experiment Setup Yes PCA is applied to further reduce the regional local features from 4096 dimensions to 2000 dimensions. The number of Gaussian distributions and the codebook size for sparse coding is set to 100 throughout our experiments unless otherwise mentioned. For all experiments, linear SVM is used as the classifier.