Zigzag Learning for Weakly Supervised Object Detection

Abstract

This paper addresses weakly supervised object detection with only image-level supervision at training stage. Previous approaches train detection models with entire images all at once, making the models prone to being trapped in sub-optimums due to the introduced false positive examples. Unlike them, we propose a zigzag learning strategy to simultaneously discover reliable object instances and prevent the model from overfitting initial seeds. Towards this goal, we first develop a criterion named mean Energy Accumulation Scores (mEAS) to automatically measure and rank localization difficulty of an image containing the target object, and accordingly learn the detector progressively by feeding examples with increasing difficulty. In this way, the model can be well prepared by training on easy examples for learning from more difficult ones and thus gain a stronger detection ability more efficiently. Furthermore, we introduce a novel masking regularization strategy over the high level convolutional feature maps to avoid overfitting initial samples. These two modules formulate a zigzag learning process, where progressive learning endeavors to discover reliable object instances, and masking regularization increases the difficulty of finding object instances properly. We achieve 47.6% mAP on PASCAL VOC 2007, surpassing the state-of-the-arts by a large margin.

WSOD2: Learning Bottom-Up and Top-Down Objectness Distillation for Weakly-Supervised Object Detection

“…We randomly initialize the weights of all new layers using Gaussian distributions with 0-mean and standard deviations 0.01 (except 0.001 for bounding box regressor), and initialize all new biases to 0. We follow a widely-used setting [2,29,30,37] to use Selective Search [33] to generate about 2, 000 proposals for each image. The whole network is end-to-end optimized using SGD with an initial learning rate of 10 −3 , weight decay of 0.0005 and momentum of 0.9.…”

“…We can observe the partial, correct, and oversized detection results for an object instance in the first, second, and third row, respectively. relieve the heavy labeling effort and reduce cost, weaklysupervised object detection paradigm has been proposed by leveraging only image-level annotations [2,30,37,38].…”

Towards Precise End-to-End Weakly Supervised Object Detection Network

“…The main task of MIL-based detectors is to learn the discriminative representation of the object instances and then select them from positive images to train a detector. Previous works on applying MIL to WSOD can be roughly categorized into multi-phase learning approach [18,4,22,38,30,42,43,41] and end-to-end learning approach [1,39,34,19,33]. End-to-end learning approaches combine CNNs and MIL into a unified network to address weakly supervised object detection task.…”

Object-Aware Instance Labeling for Weakly Supervised Object Detection

“…Using these methods, they generated pseudo ground truth boxes from localization result of OICR [20] and trained a faster R-CNN [15] model. Zhang et al [26] proposed a zigzag learning strategy, in which they developed a criterion (the Energy Accumulation Score) to automatically measure and rank localization difficulty. As the localization result of WSOD is unreliable, at first they used easy images to localize and added difficult images progressively.…”