IvaNet: Learning to jointly detect and segment objets with the help of Local Top-Down Modules
This work addresses a specific bottleneck in multi-task learning for computer vision, offering an incremental improvement over existing methods.
The paper tackles the problem of limited robustness in jointly performing object detection and semantic segmentation by introducing IvaNet, which uses local top-down modules to augment lower layers with semantic information from higher layers, achieving competitive results on PASCAL VOC and MS COCO datasets.
Driven by Convolutional Neural Networks, object detection and semantic segmentation have gained significant improvements. However, existing methods on the basis of a full top-down module have limited robustness in handling those two tasks simultaneously. To this end, we present a joint multi-task framework, termed IvaNet. Different from existing methods, our IvaNet backwards abstract semantic information from higher layers to augment lower layers using local top-down modules. The comparisons against some counterparts on the PASCAL VOC and MS COCO datasets demonstrate the functionality of IvaNet.