Local Descriptors Encoded by Fisher Vectors for Person Re-identification

By Bingpeng Ma, Yu Su, Frédéric Jurie. ECCV Workshops, 2012

371 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/a105f1ef67b4b02da38eadce8ffb4e13aa301a93

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

1. Yi Yang
2. Xiaogang Wang
3. Liang Zheng
4. Qi Tian
5. Luc Van Gool
6. Dacheng Tao
7. Tao A. Xiang
8. Wanli Ouyang

Citations in the corpus

• Person re-identification by Local Maximal Occurrence representation and metric learning (2014, methodology): “For feature representation, several effective approaches have been proposed , for example, the ensemble of local features (ELF) [10], SDALF [2], kBiCov [33], fisher vectors (LDFV) [32], salience match [46], and mid-level filter [48].”
• Scalable Person Re-identification: A Benchmark (2015, methodology): “First, it well accommodates local features, which are indicated as effective by previous works [25, 38].”
• Local Fisher Discriminant Analysis for Pedestrian Re-identification (2013, methodology): “However, the authors of eLDFV [26] also report the performance of sLDFV, a variant of eLDFV when combined with a metric learning method, which outperforms LF by 2.”
• Partial Person Re-Identification (2015, background): “Person re-identification (re-id) has been studied extensively in the past five years [5, 30, 4, 37, 10, 20, 36].”
• Unsupervised Salience Learning for Person Re-identification (2013, methodology): “On SEQ.#1, our SDC approach has better results than supervised methods, PLS and RPLM, and has comparable performance with the recently proposed eLDFV[22].”
• Part-Aligned Bilinear Representations for Person Re-identification (2018, methodology): “The early solutions of person re-identification mainly relied on hand-crafted features [20– 23], metric learning techniques [2,24–29], and probabilistic patch matching algorithms [30– 32] to handle resolution/light/view/pose changes.”
• Animal Detection From Highly Cluttered Natural Scenes Using Spatiotemporal Object Region Proposals and Patch Verification (2016, methodology): “To address this issue, we propose to encode the HOG features using FV [38] to extract high-level invariant statistical features.”
• Person Re-identification by Deep Learning Multi-scale Representations (2017, background): “Many different handcrafted person image feature descriptors [13, 74, 61, 35, 70, 38, 29] have been designed in the past decade.”
• Re-id: Hunting Attributes in the Wild (2014, methodology): “Some classic representations such as ELF [11] are fully hand designed, while others such as SDALF [9], Fisher Vectors [25] and saliency [36] are learned to a a greater or lesser extent.”
• Deep Transfer Learning for Person Re-Identification (2016, background): “Earlier works focus on either designing viewinsensitive feature representations [2], [3], or learning an effective distance metric [4], [5], or both [6].”
• A survey of approaches and trends in person re-identification (2014, positive): “Fisher vector [49] ✓ M used, single/multi-shot incorporation and the incorporation of false match rejection (novelty detection) in the matching framework.”
• Person Re-identification: Past, Present and Future (2016, methodology): “This multi-shot matching strategy is adopted by later works [97], [98].”
• Enhancing Person Re-identification in a Self-Trained Subspace (2017, methodology): “Earlier works include fisher vector based local descriptors [29], color invariant features [30] and saliency learning based methods [28].”
• Salient Color Names for Person Re-identification (2014, background): “It is a commonly used mask (or a revised mask) in person re-identification [1,18,23,20].”
• Coupled ensemble graph cuts and object verification for animal segmentation from highly cluttered videos (2015, positive): “In addition, we also encode the local gradients using Fisher vector [10] due to its effectiveness in handling low level features.”
• Dissimilarity-based people re-identification and search for intelligent video surveillance (2013, methodology): “Maximally Stable Colour Regions (MSCR) [51] are used in [28, 45, 88].”
• Hierarchical Gaussian Descriptor for Person Re-identification (2016, methodology): “Other meta descriptors: We compare the GOG descriptor with other meta-descriptors: Heterogeneous AutoSimilarities of Characteristics (HASC) [7], Local Descriptors encoded by Fisher Vector (LDFV) [28], Second-order Average Pooling (2AvgP) [9] and GOLD [37].”