# jemdoc: showsource = Learning Collections of Part Models for Object Recognition\n\n [http://web.engr.illinois.edu/~iendres2/ Ian Endres], [http://web.engr.illinois.edu/~kjshih2/ Kevin J Shih], Johnston Jiaa, [http://www.cs.uiuc.edu/~dhoiem/ Derek Hoiem] == Download A release version for the code is available in the following link. Some changes have been made to the classifier code to improve stability, so the performance will be similar, but will not exactly match the numbers reported in the paper.\n \[[http://web.engr.illinois.edu/~kjshih2/code/bcp_release_1.0.tar.gz source]\] == Results In the paper, the shape rescoring features erroneously included scores from color and texture classifiers in addition to the region shape classifier. The detection results for Boat, Car, and Person in table 3 used parts trained with an old version of consistency, leading to significantly different results. To ensure that the correct detectors were used, all models and classifiers were retrained from scratch, leading to some variance in the results. Below, *shape* includes features for aspect ratio, scale, HOG shape classifier, and Object Proposal rank. *Reloc* indicates the bounding box reestimation described in the paper. ~~~ {}{table}{Results} Category | aeroplane | bicycle | bird | boat| bottle | bus | car | cat | chair | cow | diningtable | dog | horse | motorbike | person | pottedplant | sheep | sofa | train | tvmonitor | meanAP|| Reported performance | .44340 |.35240 |.09700 |.101 |.15120 |.44590| .3207 |.3534 |.0438 |.17490 |.14960 |.27620 |.36240 |.42080 |.3 |.0501| .1377| .1877| .34390| .2859 |.2499|| Parts \+ Shape \+ Reloc | .44150 |.3728 |.0927| .0504| .1672 |.462| .2851| .3374 |.0751| .1695 | .1869| .256| .3628| .4211 |.2616 |.0457 |.1663 |.1829 | .3146 |.2886 |.2469 || Parts \+ Shape \+ Texture \+ Color \+ Reloc | .4465 |.3753 |.0968 |.0452| .1603 |.4599 |.2843 |.3426 |.0799| .1656| .1836| .2569| .3627| .4293 |.2621| .0586 |.1717| .1855 |.314 |.29 |.2484 || ~~~ == References *Learning Collections of Part Models for Object Recognition*\n I Endres, KJ Shih, J Jiaa, D Hoiem\n CVPR 2013\n \[[http://www.cv-foundation.org/openaccess/content_cvpr_2013/papers/Endres_Learning_Collections_of_2013_CVPR_paper.pdf pdf]\]