DeepProposal: Hunting Objects by Cascading Deep Convolutional Layers

Abstract

In this paper we evaluate the quality of the activation layers of a convolutional neural network (CNN) for the generation of object proposals. We generate hypotheses in a sliding-window fashion over different activation layers and show that the final convolutional layers can find the object of interest with high recall but poor localization due to the coarseness of the feature maps. Instead, the first layers of the network can better localize the object of interest but with a reduced recall. Based on this observation we design a method for proposing object locations that is based on CNN features and that combines the best of both worlds. We build an inverse cascade that, going from the final to the initial convolutional layers of the CNN, selects the most promising object locations and refines their boxes in a coarse-to-fine manner. The method is efficient, because i) it uses the same features extracted for detection, ii) it aggregates features using integral images, and iii) it avoids a dense evaluation of the proposals due to the inverse coarse-to-fine cascade. The method is also accurate; it outperforms most of the previously proposed object proposals approaches and when plugged into a CNN-based detector produces state-of-theart detection performance.

Harvesting Visual Objects from Internet Images via Deep-Learning-Based Objectness Assessment

“…For example, network trained on COB proposals is denoted as "COB+OA". We compare COB+OA against nine state-of-the-art object proposal generators, including Objectness Measure [4], Contour Box [31], Selective Search [47], Edge Box [63], MCG [5], COB [34], Deep Box [24], Deep Proposal [16], and RPN [38]. The last four are the latest deep learning based methods.…”

“…Instead of directly searching for optimal windows, we use off-the-shelves object proposal generators for proposal box generation. Several methods are tried, including Selective Search [47], Edge Box [63], MCG [5], RPN from Faster R-CNN [38], and Deep Proposal [16]. In this work, we prefer fast proposal generation methods with good precision so that later our assessment network can indeed find optimal object proposals.…”

Adaptive video object proposals by a context-aware model

“…There is no exception in the task of object proposals [19,20,30]. Zhang et al [64] leveraged a Convolutional-Neural-Network model to generate location proposals of salient objects.…”

Region Proposal by Guided Anchoring

“…MetaAnchor [32] introduces meta-learning to anchor generation. There have been attempts [8,9,23,31,33,34,1,2] that apply cascade architecture to reject easy samples at early layers or stages, and regress bounding boxes iteratively for progressive refinement. Compared to two-stage approaches, the single-stage pipeline skips object proposal generation and predicts bounding boxes and class scores in one evaluation.…”