Collaborative Learning-based Network for Weakly Supervised Remote Sensing Object Detection

Existing object detection algorithms rely excessively on instance-level labels, which are both time-consuming and expensive. In particular, for remote sensing images (RSI) with small dense objects, the labeling cost is much higher than that of general images. Moreover, propagation process labels over noisy channel results in blurred information. To address problem obtaining RSI we aim to propose a collaborative learning-based network weakly supervised (CLN-RSOD). Compared state-of-the-art, proposed model combines advantages two sub-networks by jointly training. This improves model's capability thereby enhances effect multi-object RSI. employ mask-based proposal refinement algorithm (MPR-RS) optimize candidate boxes. addition, according data distribution characteristics RSI, introduce new joint pooling module (JPM) CLN-RSOD enhance backbone network's characterization Finally, experimental public datasets illustrate learning method superior other methods demonstrates effectiveness CLN-RSOD.