Class-agnostic counting is usually phrased as a matching problem between a user-defined exemplar patch and a query image. The count is derived based on the number of objects similar to the exemplar patch. However, defining a target class using only exemplar patches inevitably miscounts unintended objects that are visually alike to the exemplar. In this paper, we propose to include negative exemplars that define what not to count in order to disentangle visually similar negatives, leading to a more discriminative definition of the target object. This allows the model to calibrate its notion of what is similar based on both positive and negative exemplars. We outperformed state-of-the-art by adding a few negative exemplars, improving the MAE by 4.06 points or 18.40% improvement. Moreover, our model can be incorporated into a semi-automatic labeling tool to simplify the job of the annotator.

[1]X.Wu,Y.Zheng,H.Ye,W.Hu,J.Yang,andL.He,“Adaptivescenariodiscoveryforcrowdcounting,”inICASSP 2019-2019 IEEE International Conference on Acoustics, Speech and Signal Processing(ICASSP), pp. 2382–2386, IEEE, 2019.
[2]N. N. Bhookya, R. Malmathanraj, and P. Palanisamy, “Yield estimation of chilli crop using imageprocessing techniques,” in2020 6th International Conference on Advanced Computing and Commu-nication Systems (ICACCS), pp. 200–204, IEEE, 2020.
[3]M. L. Mekhalfi, C. Nicolò, I. Ianniello, F. Calamita, R. Goller, M. Barazzuol, and F. Melgani, “Visionsystemforautomaticon-treekiwifruitcountingandyieldestimation,”Sensors,vol.20,no.15,p.4214,2020.
[4]U.-O. Dorj, M. Lee, and S.-s. Yun, “An yield estimation in citrus orchards via fruit detection andcounting using image processing,”Computers and Electronics in Agriculture, vol. 140, pp. 103–112,2017.
[5]Y. Yang, G. Li, Z. Wu, L. Su, Q. Huang, and N. Sebe, “Reverse perspective network for perspective-awareobjectcounting,”inProceedings of the IEEE/CVF Conference on Computer Vision and PatternRecognition, pp. 4374–4383, 2020.
[6]X. Jiang, L. Zhang, M. Xu, T. Zhang, P. Lv, B. Zhou, X. Yang, and Y. Pang, “Attention scalingfor crowd counting,” inProceedings of the IEEE/CVF Conference on Computer Vision and PatternRecognition, pp. 4706–4715, 2020.
[7]Z. Shi, P. Mettes, and C. G. Snoek, “Counting with focus for free,” inProceedings of the IEEE/CVFInternational Conference on Computer Vision, pp. 4200–4209, 2019.
[8]C.Liu, X.Weng, andY.Mu, “Recurrentattentivezoomingforjointcrowdcountingandpreciselocal-ization,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition,pp. 1217–1226, 2019.
[9]J. Huang, G. Ding, Y. Guo, D. Yang, S. Wang, T. Wang, and Y. Zhang, “Drone-based car counting viadensity map learning,” in2020 IEEE International Conference on Visual Communications and ImageProcessing (VCIP), pp. 239–242, IEEE, 2020.
[10]L. Ciampi, G. Amato, F. Falchi, C. Gennaro, and F. Rabitti, “Counting vehicles with cameras.,” inSEBD, 2018.
[11]Y. Zhong, J. Gao, Q. Lei, and Y. Zhou, “A vision-based counting and recognition system for flyinginsects in intelligent agriculture,”Sensors, vol. 18, no. 5, p. 1489, 2018.
[12]E. A. Lins, J. P. M. Rodriguez, S. I. Scoloski, J. Pivato, M. B. Lima, J. M. C. Fernandes, P. R. V.daSilvaPereira,D.Lau,andR.Rieder,“Amethodforcountingandclassifyingaphidsusingcomputervision,”Computers and Electronics in Agriculture, vol. 169, p. 105200, 2020.
[13]E. C. Tetila, B. B. Machado, G. V. Menezes, N. A. de Souza Belete, G. Astolfi, and H. Pistori, “Adeep-learningapproachforautomaticcountingofsoybeaninsectpests,”IEEEGeoscienceandRemoteSensing Letters, vol. 17, no. 10, pp. 1837–1841, 2019.
[14]B. Arshad, J. Barthelemy, E. Pilton, and P. Perez, “Where is my deer?-wildlife tracking and countingvia edge computing and deep learning,” in2020 IEEE Sensors, pp. 1–4, IEEE, 2020.
[15]B. Xu, W. Wang, G. Falzon, P. Kwan, L. Guo, G. Chen, A. Tait, and D. Schneider, “Automated cattlecounting using mask r-cnn in quadcopter vision system,”Computers and Electronics in Agriculture,vol. 171, p. 105300, 2020.
[16]V. Ranjan, U. Sharma, T. Nguyen, and M. Hoai, “Learning to count everything,” inProceedings ofthe Conference on Computer Vision and Pattern Recognition (CVPR), 2021.
[17]E. Lu, W. Xie, and A. Zisserman, “Class-agnostic counting,” inAsian conference on computer vision,pp. 669–684, Springer, 2018.
[18]S.-D. Yang, H.-T. Su, W. H. Hsu, and W.-C. Chen, “Class-agnostic few-shot object counting,” inProceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, pp. 870–878,2021.
[19]Q. Fan, W. Zhuo, C.-K. Tang, and Y.-W. Tai, “Few-shot object detection with attention-rpn and multi-relation detector,” inProceedings of the IEEE/CVF Conference on Computer Vision and PatternRecognition, pp. 4013–4022, 2020.
[20]B. Kang, Z. Liu, X. Wang, F. Yu, J. Feng, and T. Darrell, “Few-shot object detection via fea-ture reweighting,” inProceedings of the IEEE/CVF International Conference on Computer Vision,pp. 8420–8429, 2019.
[21]P. Chattopadhyay, R. Vedantam, R. R. Selvaraju, D. Batra, and D. Parikh, “Counting everyday ob-jects in everyday scenes,” inProceedings of the IEEE Conference on Computer Vision and PatternRecognition, pp. 1135–1144, 2017.
[22]M.-R.Hsieh,Y.-L.Lin,andW.H.Hsu,“Drone-basedobjectcountingbyspatiallyregularizedregionalproposalnetwork,”inProceedingsoftheIEEEinternationalconferenceoncomputervision,pp.4145–4153, 2017.
[23]V. Lempitsky and A. Zisserman, “Learning to count objects in images,”Advances in neural informa-tion processing systems, vol. 23, pp. 1324–1332, 2010.
[24]Y. Zhang, D. Zhou, S. Chen, S. Gao, and Y. Ma, “Single-image crowd counting via multi-columnconvolutionalneuralnetwork,”inProceedingsoftheIEEEconferenceoncomputervisionandpatternrecognition, pp. 589–597, 2016.
[25]C. Finn, P. Abbeel, and S. Levine, “Model-agnostic meta-learning for fast adaptation of deep net-works,” inInternational Conference on Machine Learning, pp. 1126–1135, PMLR, 2017.
[26]G. Koch, R. Zemel, R. Salakhutdinov,et al., “Siamese neural networks for one-shot image recogni-tion,” inICML deep learning workshop, vol. 2, Lille, 2015.[27]B. M. Lake, R. Salakhutdinov, and J. B. Tenenbaum, “Human-level concept learning through proba-bilistic program induction,”Science, vol. 350, no. 6266, pp. 1332–1338, 2015.
[28]S. Ravi and H. Larochelle, “Optimization as a model for few-shot learning,” 2016.
[29]A.Santoro,S.Bartunov,M.Botvinick,D.Wierstra,andT.Lillicrap,“One-shotlearningwithmemory-augmented neural networks,”arXiv preprint arXiv:1605.06065, 2016.
[30]O. Vinyals, C. Blundell, T. Lillicrap, D. Wierstra,et al., “Matching networks for one shot learning,”Advances in neural information processing systems, vol. 29, pp. 3630–3638, 2016.
[31]T. Chen, S. Kornblith, M. Norouzi, and G. Hinton, “A simple framework for contrastive learningof visual representations,” inInternational conference on machine learning, pp. 1597–1607, PMLR,2020.
[32]J.-B. Grill, F. Strub, F. Altché, C. Tallec, P. H. Richemond, E. Buchatskaya, C. Doersch, B. A. Pires,Z. D. Guo, M. G. Azar,et al., “Bootstrap your own latent: A new approach to self-supervised learn-ing,”arXiv preprint arXiv:2006.07733, 2020.
[33]K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” inProceedingsof the IEEE conference on computer vision and pattern recognition, pp. 770–778, 2016.
[34]J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei, “Imagenet: A large-scale hierarchicalimage database,” in2009 IEEE conference on computer vision and pattern recognition, pp. 248–255,Ieee, 2009.
[35]B. Settles, “Active learning literature survey,” 2009.