CoChat

Ensemble of exemplar-SVMs for object detection and beyond

Open access
Malisiewicz, Tomasz, Gupta, Abhinav, Efros, Alexei A. · 2011 · 2011 International Conference on Computer Vision

The DOI matches its Crossref record; 23 of the 25 references it cites were found in a registry or an index.

Abstract

This paper proposes a conceptually simple but surprisingly powerful method which combines the effectiveness of a discriminative object detector with the explicit correspondence offered by a nearest-neighbor approach. The method is based on training a separate linear SVM classifier for every exemplar in the training set. Each of these Exemplar-SVMs is thus defined by a single positive instance and millions of negatives. While each detector is quite specific to its exemplar, we empirically observe that an ensemble of such Exemplar-SVMs offers surprisingly good generalization. Our performance on the PASCAL VOC detection task is on par with the much more complex latent part-based model of Felzenszwalb et al., at only a modest computational cost increase. But the central benefit of our approach is that it creates an explicit association between each detection and a single training exemplar. Because most detections show good alignment to their associated exemplar, it is possible to transfer any available exemplar meta-data (segmentation, geometric structure, 3D model, etc.) directly onto the detections, which can then be used as part of overall scene understanding.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1109/iccv.2011.6126229

Cites

23 of 23 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 Object Category Detector vs. Ensemble of Exemplar Detectors. Output of a typical object detector is just a bounding box and a category label (left). But our ensemble of ExemplarSVMs is able to associate each detection with a visually similar training exemplar (right), allowing for direct transfer of meta-data such as segmentation, geometry, even a 3 D model (bottom).
Figure 1 Object Category Detector vs. Ensemble of Exemplar Detectors. Output of a typical object detector is just a bounding box and a category label (left). But our ensemble of ExemplarSVMs is able to associate each detection with a visually similar training exemplar (right), allowing for direct transfer of meta-data such as segmentation, geometry, even a 3 D model (bottom).CC-BY · From the PDF
Figure 2 Category SVM vs. Exemplar-SVMs . Instead of training a single per-category classifier, we train a separate linear SVM classifier for each exemplar in our dataset with a single positive example and millions of negative windows. Negatives come from images not containing any instances of the exemplar's category.
Figure 2 Category SVM vs. Exemplar-SVMs . Instead of training a single per-category classifier, we train a separate linear SVM classifier for each exemplar in our dataset with a single positive example and millions of negative windows. Negatives come from images not containing any instances of the exemplar's category.CC-BY · From the PDF
Figure 3 Comparison. Given a bicycle training sample from PASCAL (represented with a HOG weight vector w ), we show the top 6 matches from the PASCAL test-set using three methods. Row 1: naive nearest neighbor (using raw normalized HOG). Row 2: Trained Exemplar-SVM (notice how w focuses on bike-specific edges). Row 3: Learned distance function - an Exemplar-SVM but trained in the 'distance-to-exemplar' vector space, with the exemplar being placed at the origin (loosely corresponding to [10, 14]).
Figure 3 Comparison. Given a bicycle training sample from PASCAL (represented with a HOG weight vector w ), we show the top 6 matches from the PASCAL test-set using three methods. Row 1: naive nearest neighbor (using raw normalized HOG). Row 2: Trained Exemplar-SVM (notice how w focuses on bike-specific edges). Row 3: Learned distance function - an Exemplar-SVM but trained in the 'distance-to-exemplar' vector space, with the exemplar being placed at the origin (loosely corresponding to [10, 14]).CC-BY · From the PDF
Figure 4 Exemplar-SVMs. A few 'train' exemplars with their top detections on the PASCAL VOC test-set. Note that each exemplar's HOG has its own dimensions. Note also how each detector is specific not just to the train's orientation, but even to the type of train.
Figure 4 Exemplar-SVMs. A few 'train' exemplars with their top detections on the PASCAL VOC test-set. Note that each exemplar's HOG has its own dimensions. Note also how each detector is specific not just to the train's orientation, but even to the type of train.CC-BY · From the PDF
Figure 5 Exemplar-SVM calibration . The calibration step rescales the SVM scores but does not affect the ordering of the matches, allowing us to compare the outputs of multiple independently-trained Exemplar-SVMs.
Figure 5 Exemplar-SVM calibration . The calibration step rescales the SVM scores but does not affect the ordering of the matches, allowing us to compare the outputs of multiple independently-trained Exemplar-SVMs.CC-BY · From the PDF
Figure 6 Object Detection and Appearance Transfer. Each example shows a detection from our ensemble of Exemplar-SVMs along with the appearance transferred directly from the source exemplar, to demonstrate the high quality of visual alignment. Bottom row shows object category detection failures.
Figure 6 Object Detection and Appearance Transfer. Each example shows a detection from our ensemble of Exemplar-SVMs along with the appearance transferred directly from the source exemplar, to demonstrate the high quality of visual alignment. Bottom row shows object category detection failures.CC-BY · From the PDF
Figure 7 Segmentation Transfer. Object segmentations are transferred from the exemplar directly onto the detection window.
Figure 7 Segmentation Transfer. Object segmentations are transferred from the exemplar directly onto the detection window.CC-BY · From the PDF
Figure 8 Qualitative Geometry Transfer. We transfer geometric labeling from bus exemplars onto corresponding detections.
Figure 8 Qualitative Geometry Transfer. We transfer geometric labeling from bus exemplars onto corresponding detections.CC-BY · From the PDF
Figure 9 3D Model transfer. In each of these 3 examples, the green box in the top image shows the detection window, and the bottom shows the automatically transferred 3 D model.
Figure 9 3D Model transfer. In each of these 3 examples, the green box in the top image shows the detection window, and the bottom shows the automatically transferred 3 D model.CC-BY · From the PDF
Image 10
CC-BY · From the PDF

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 23 references

Research works printed but not linked: 1

  1. A. Frome and J. Malik. Image retrieval and recognition using local distance functions. NIPS, 2006. 2, 3, 4

Cited by

Items 301–400 of 873

Metadata

Authors
Malisiewicz, T., Gupta, A., & Efros, A. A.
Affiliations
Carnegie Mellon University (US)
Venue
2011 International Conference on Computer Vision
Publisher
IEEE
Published
2011-11
Pages
89-96
OpenAlex
W1989684337
Cited by
865
References
25
Citations by year
2026: 4 · 2025: 18 · 2024: 16 · 2023: 29 · 2022: 54 · 2021: 52 · 2020: 49 · 2019: 45 · 2018: 65 · 2017: 78 · 2016: 105 · 2015: 148 · 2014: 97 · 2013: 65 · 2012: 36
FWCI
48.23
Type
conference paper
Language
en
Open access
Licence
cc-by
Locations
  • IEEE/CVF International Conference on Computer Vision (ICCV) (publishedVersion)
  • Figshare (submittedVersion, cc-by)
  • KiltHub Repository (submittedVersion)
  • KiltHub Repository
Funders
  • National Science Foundation (0905402)
  • Multidisciplinary University Research Initiative
Topics
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Image and Video Retrieval Techniques
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Visual Attention and Saliency Detection
Keywords
object detection, linear SVM classifier, Pascal VOC, scene understanding, part-based models, generalization

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.