Discriminatively Activated Sparselets

Ross Girshick, Hyun Oh Song, Trevor Darrell
Proceedings of the 30th International Conference on Machine Learning, PMLR 28(1):196-204, 2013.

Abstract

Shared representations are highly appealing due to their potential for gains in computational and statistical efficiency. Compressing a shared representation leads to greater computational savings, but at the same time can severely decrease performance on a target task. Recently, sparselets (Song et al., 2012) were introduced as a new shared intermediate representation for multiclass object detection with deformable part models (Felzenszwalb et al., 2010a), showing significant speedup factors, but with a large decrease in task performance. In this paper we describe a new training framework that learns which sparselets to activate in order to optimize a discriminative objective, leading to larger speedup factors with no decrease in task performance. We first reformulate sparselets in a general structured output prediction framework, then analyze when sparselets lead to computational efficiency gains, and lastly show experimental results on object detection and image classification tasks. Our experimental results demonstrate that discriminative activation substantially outperforms the previous reconstructive approach which, together with our structured output prediction formulation, make sparselets broadly applicable and significantly more effective.

Cite this Paper


BibTeX
@InProceedings{pmlr-v28-girshick13, title = {Discriminatively Activated Sparselets}, author = {Girshick, Ross and Song, Hyun Oh and Darrell, Trevor}, booktitle = {Proceedings of the 30th International Conference on Machine Learning}, pages = {196--204}, year = {2013}, editor = {Dasgupta, Sanjoy and McAllester, David}, volume = {28}, number = {1}, series = {Proceedings of Machine Learning Research}, address = {Atlanta, Georgia, USA}, month = {17--19 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v28/girshick13.pdf}, url = {https://proceedings.mlr.press/v28/girshick13.html}, abstract = {Shared representations are highly appealing due to their potential for gains in computational and statistical efficiency. Compressing a shared representation leads to greater computational savings, but at the same time can severely decrease performance on a target task. Recently, sparselets (Song et al., 2012) were introduced as a new shared intermediate representation for multiclass object detection with deformable part models (Felzenszwalb et al., 2010a), showing significant speedup factors, but with a large decrease in task performance. In this paper we describe a new training framework that learns which sparselets to activate in order to optimize a discriminative objective, leading to larger speedup factors with no decrease in task performance. We first reformulate sparselets in a general structured output prediction framework, then analyze when sparselets lead to computational efficiency gains, and lastly show experimental results on object detection and image classification tasks. Our experimental results demonstrate that discriminative activation substantially outperforms the previous reconstructive approach which, together with our structured output prediction formulation, make sparselets broadly applicable and significantly more effective.} }
Endnote
%0 Conference Paper %T Discriminatively Activated Sparselets %A Ross Girshick %A Hyun Oh Song %A Trevor Darrell %B Proceedings of the 30th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2013 %E Sanjoy Dasgupta %E David McAllester %F pmlr-v28-girshick13 %I PMLR %P 196--204 %U https://proceedings.mlr.press/v28/girshick13.html %V 28 %N 1 %X Shared representations are highly appealing due to their potential for gains in computational and statistical efficiency. Compressing a shared representation leads to greater computational savings, but at the same time can severely decrease performance on a target task. Recently, sparselets (Song et al., 2012) were introduced as a new shared intermediate representation for multiclass object detection with deformable part models (Felzenszwalb et al., 2010a), showing significant speedup factors, but with a large decrease in task performance. In this paper we describe a new training framework that learns which sparselets to activate in order to optimize a discriminative objective, leading to larger speedup factors with no decrease in task performance. We first reformulate sparselets in a general structured output prediction framework, then analyze when sparselets lead to computational efficiency gains, and lastly show experimental results on object detection and image classification tasks. Our experimental results demonstrate that discriminative activation substantially outperforms the previous reconstructive approach which, together with our structured output prediction formulation, make sparselets broadly applicable and significantly more effective.
RIS
TY - CPAPER TI - Discriminatively Activated Sparselets AU - Ross Girshick AU - Hyun Oh Song AU - Trevor Darrell BT - Proceedings of the 30th International Conference on Machine Learning DA - 2013/02/13 ED - Sanjoy Dasgupta ED - David McAllester ID - pmlr-v28-girshick13 PB - PMLR DP - Proceedings of Machine Learning Research VL - 28 IS - 1 SP - 196 EP - 204 L1 - http://proceedings.mlr.press/v28/girshick13.pdf UR - https://proceedings.mlr.press/v28/girshick13.html AB - Shared representations are highly appealing due to their potential for gains in computational and statistical efficiency. Compressing a shared representation leads to greater computational savings, but at the same time can severely decrease performance on a target task. Recently, sparselets (Song et al., 2012) were introduced as a new shared intermediate representation for multiclass object detection with deformable part models (Felzenszwalb et al., 2010a), showing significant speedup factors, but with a large decrease in task performance. In this paper we describe a new training framework that learns which sparselets to activate in order to optimize a discriminative objective, leading to larger speedup factors with no decrease in task performance. We first reformulate sparselets in a general structured output prediction framework, then analyze when sparselets lead to computational efficiency gains, and lastly show experimental results on object detection and image classification tasks. Our experimental results demonstrate that discriminative activation substantially outperforms the previous reconstructive approach which, together with our structured output prediction formulation, make sparselets broadly applicable and significantly more effective. ER -
APA
Girshick, R., Song, H.O. & Darrell, T.. (2013). Discriminatively Activated Sparselets. Proceedings of the 30th International Conference on Machine Learning, in Proceedings of Machine Learning Research 28(1):196-204 Available from https://proceedings.mlr.press/v28/girshick13.html.

Related Material