Training Deformable Object Models for Human Detection Based on Alignment and Clustering

Abstract

Abstract. We propose a clustering method that considers non-rigid alignment of samples. The motivation for such a clustering is training of object detectors that consist of multiple mixture components. In particular, we consider the deformable part model (DPM) of Felzenszwalb et al., where each mixture component includes a learned deformation model. We show that alignment based clustering distributes the data better to the mixture components of the DPM than previous methods. Moreover, the alignment helps the non-convex optimization of the DPM find a consistent placement of its parts and, thus, learn more accurate part filters.

Exploring Deep Learning for Complex Trait Genomic Prediction in Polyploid Outcrossing Species

“…Numerous successful applications of DL in classification contexts have been published, e.g. pattern recognition (Drayer and Brox, 2014;Liang and Hu, 2015;Işin et al, 2016;Badrinarayanan et al, 2017) and natural language processing (NLP) (Deng and Liu, 2018). The DL implementation in regression tasks is less abundant and the benefit of using these methods remains uncertain (Bellot et al, 2018;Montesinos-López et al, 2018a;Azodi et al, 2019).…”

Discovering object aspects from video

“…Several methods [5,6,7,8,9,10,17,18,19] discover aspects implicitly, in order to train specialised classifiers for each of them (components of a mixture model). Some of these works [5,6,7,8] cluster HOG descriptors extracted from bounding boxes in the training images (manually annotated). Others [9,10] use exemplar SVMs [20] as a similarity measure between bounding boxes to drive the clustering.…”

“…While some recent methods discover aspects from still images [5,6,7,8,9,10], they all require manual annotations of the object's location (e.g. bounding boxes).…”

“…Our exploration relies on a new protocol for evaluating aspect discovery directly. In contrast, previous works evaluate aspect discovery indirectly, typically by measuring its impact on object detection performance [5,6,7,8]. For this, we collected a large dataset sourced from videos of two different classes, car and tiger (for a total of 2664 video shots, sec.…”

Learning detector ensembles for object detection

“…Object detection is a fundamental problem in computer vision and has received much attention in past few decades [3, 7-10, 12, 14, 37, 38]. Many approaches have been widely adopted to address appearance variations for object detection, including sub-category learning [18][19][20]39], deformable part models [10,26,[39][40][41][42][43] and occlusion handling [2, 22-25, 31, 33, 44, 45]. Below we summarize some works which are most related to the contributions of this thesis.…”