Evaluation of GIST descriptors for web-scale image search

By Matthijs Douze, Herv'e J'egou, Harsimrat Sandhawalia, Laurent Amsaleg, Cordelia Schmid. 2009

460 citations recorded by the source corpus.

Subfields: Aerospace Engineering, 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 Aerospace Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

Microsoft COCO: Common Objects in Context (2014, methodology): “We took care to minimize the chance of near-duplicate images existing across splits by explicitly removing near duplicates (detected with [43]) and grouping images by photographer and date taken.”
Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation (2013, methodology): “Following [13], we computed GIST descriptors on warped 32 × 32 pixel versions of all ILSVRC 2012 trainval and PASCAL 2007 test images.”
Product Quantization for Nearest Neighbor Search (2011, methodology): “When no supervision is used, this mapping is learned such that the neighborhood in the embedded space deﬁned by the Hamming distance reﬂects the neighborhood in the Euclidean space of the original features.”
Remote Sensing Image Scene Classification: Benchmark and State of the Art (2017, neutral): “popularly used for scene representation [111-113].”
PlaNet - Photo Geolocation with Convolutional Neural Networks (2016, methodology): “ Bottom: Ground truth location (yellow), human guess (green), PlaNet guess (blue). Method Holidays mAP BoVW 57.2 [24] Hamming Embedding 77.5 [24] Fine Vocabulary 74.9 [34] ASMK+MA 82.2 [47] GIST 37.6 [10] Overfeat 64.2 [39] Overfeat+aug+ss 84.3 [39] AlexNet+LM Retraining 79.3 [3] PlaNet (this work) ”
Emerging Properties in Self-Supervised Vision Transformers (2021, methodology): “We report the mAP performance in copy detection on Copydays “strong” subset [21].”
NetVLAD: CNN Architecture for Weakly Supervised Place Recognition (2015, methodology): “some form of distinctiveness of local features [2], [9], [12], [63], [64], [65], [66]), and (ii) learning on top of shallow hand-engineered descriptors that cannot be finetuned for the target task [2], [6], [7], [48], [67].”
MultiGrain: a unified image embedding for classes and instances (2019, methodology): “We also report the performance of our network in a copy detection setting, indicating the mean average precision on the “strong” subset of the INRIA Copydays dataset [8].”
Are Labels Required for Improving Adversarial Robustness? (2019, methodology): “to images obtained from web queries matching the CIFAR-10 classes, and filter out near duplicates of the CIFAR-10 test set using GIST features [34, 15].”
Contextual Enrichment of Remote-Sensed Events with Social Media Streams (2016, methodology): “Our evaluation is based on the INRIA CopyDays dataset[4] which contains 157 original images and includes three transformations (1) 16% resizing of the image surface plus diﬀerent JPEQ-Quality compressions (2) cropping of the image surface between 5% to 80% as well as (3) strong transformations such…”
Do deep features generalize from everyday objects to remote sensing and aerial scenes domains? (2015, methodology): “GIST implementation is the one used in [13] with the parameters discussed therein.”
Improving Bag-of-Features for Large Scale Image Search (2010, methodology): “2This approach has also been applied to GIST descriptors in [18].”
Large-scale image retrieval with compressed Fisher vectors (2010, background): “While GIST descriptors can be very effective to retrieve similar objects and scenes which are well aligned, they cannot cope with wide variations in rotation, scaling or viewpoint [7].”
Aggregating local descriptors into a compact image representation (2010, background): “Some authors address the efficiency and memory constraints by using GIST descriptors [17], and by converting them to compact binary vectors [5, 11, 24, 27].”