Imperial College London


Faculty of EngineeringDepartment of Electrical and Electronic Engineering

Reader in Computer Vision



+44 (0)20 7594 6220k.mikolajczyk




Electrical EngineeringSouth Kensington Campus






BibTex format

author = {Koniusz, P and Yan, F and Mikolajczyk, K},
doi = {10.1016/j.cviu.2012.10.010},
journal = {Computer Vision and Image Understanding},
pages = {479--492},
title = {Comparison of mid-level feature coding approaches and pooling strategies in visual concept detection},
url = {},
volume = {117},
year = {2013}

RIS format (EndNote, RefMan)

AB - Bag-of-Words lies at a heart of modern object category recognition systems. After descriptors are extracted from images, they are expressed as vectors representing visual word content, referred to as mid-level features. In this paper, we review a number of techniques for generating mid-level features, including two variants of Soft Assignment, Locality-constrained Linear Coding, and Sparse Coding. We also isolate the underlying properties that affect their performance. Moreover, we investigate various pooling methods that aggregate mid-level features into vectors representing images. Average pooling, Max-pooling, and a family of likelihood inspired pooling strategies are scrutinised. We demonstrate how both coding schemes and pooling methods interact with each other. We generalise the investigated pooling methods to account for the descriptor interdependence and introduce an intuitive concept of improved pooling. We also propose a coding-related improvement to increase its speed. Lastly, state-of-the-art performance in classification is demonstrated on Caltech101, Flower17, and ImageCLEF11 datasets. © 2012 Elsevier Inc. All rights reserved.
AU - Koniusz,P
AU - Yan,F
AU - Mikolajczyk,K
DO - 10.1016/j.cviu.2012.10.010
EP - 492
PY - 2013///
SN - 1077-3142
SP - 479
TI - Comparison of mid-level feature coding approaches and pooling strategies in visual concept detection
T2 - Computer Vision and Image Understanding
UR -
VL - 117
ER -