Learning To Count Objects in Images

By Victor S. Lempitsky, Andrew Zisserman. 2010

1,352 citations recorded by the source corpus.

Subfields: 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 Computer Vision and Pattern Recognition

Citations in the corpus

DeepSetNet: Predicting Sets with Deep Neural Networks (2016, background): “It is important to bear in mind that unlike many existing approaches that learn to count [1, 5, 12, 20, 24, 35, 50, 51], our main goal is not object counting.”
End-to-End People Detection in Crowded Scenes (2015, background): “es and summarize results in each experiment with average precision (AP) and equal error rate (EER) in Fig. 5. We also analyze how well each model predicts the total count of people in an image. As in [13], we measure count error by computing the average absolute difference between the number of pred”
Finely-grained annotated datasets for image-based plant phenotyping (2016, methodology): “Learning-based counting techniques could help here (Lempitsky and Zisserman, 2010; Fiaschi et al., 2013).”
Domain Adaptation for Traffic Density Estimation (2021, methodology): “Instead, Counting by regression (Lempitsky and Zisserman, 2010) is a supervised learning approach that tries to establish a direct mapping (linear or not) from the image features to the number of objects present in the scene or a corresponding density map (i.”
Representation Learning by Learning to Count (2017, background): “The counting prediction can be provided as an object density map [15, 39, 23] or simply as the number of counted objects [5, 6].”
Attend, Infer, Repeat: Fast Scene Understanding with Generative Models (2016, background): “By its nature AIR is also related to the following problems: counting [14, 27], pondering [2], and gradient estimation through renderers [15].”
Learning to Count Objects in Natural Images for Visual Question Answering (2018, methodology): “More traditional approaches based on Lempitsky & Zisserman (2010) learn to produce a target density map, from which a count is computed by integrating over it.”
End-to-End Instance Segmentation and Counting with Recurrent Attention (2016, methodology): “Previous work on object counting in images has mainly focused on crowds of pedestrians and biological cells [8].”
Inference Compilation and Universal Probabilistic Programming (2016, methodology): “Figure 2 shows one kind of application such an efficient inference engine can do: object counting (Lempitsky and Zisserman, 2010) and localization.”
Loss Functions for Multiset Prediction (2017, background): “[12, 28, 15], model the counting problem as density estimation over image space, and assume that each object is annotated with a dot specifying its location.”
Iterative Crowd Counting (2018, background): “Lempitsky & Zisserman [10] learned a linear mapping between the crowd images and the corresponding ground truth density maps.”
Context-Aware Crowd Counting (2018, methodology): “The regressors can be based on Random Forests [18], Gaussian Processes [7], or more recently Deep Nets [41, 42, 26, 31, 40, 36, 32, 24, 19, 30, 33, 22, 15, 28, 5], with most state-of-the-art approaches now relying on the latter.”
Single-Image Crowd Counting via Multi-Column Convolutional Neural Network (2016, methodology): “[17] adopts dense SIFT features on randomly selected patches and the MESA distance to learn a density regression model.”
CSRNet: Dilated Convolutional Neural Networks for Understanding the Highly Congested Scenes (2018, methodology): “We compare our approach with the previous state-ofthe-art methods [47, 32, 20, 24].”
Counting Vehicles with Cameras (2018, methodology): “Finally, counting by density estimation is a supervised technique that extends in some way counting by regression approach, introduced in [8].”
Learning to Detect Partially Overlapping Instances (2013, methodology): “In Table 1 we show the performance of our method and its components for the case of 32 + 32 training+validation images (same splits as in [15]).”
Density-aware person detection and tracking in crowds (2011, methodology): “We learn w from k training samples by minimizing the regularized MESA distance D introduced in [18] as”
COUNT Forest: CO-Voting Uncertain Number of Targets Using Random Forest for Crowd Density Estimation (2015, methodology): “learn a linear transformation of the feature representation f(x) that approximates the density function at each pixel D(x) = ω f(x) [21].”