Unseen Object Segmentation in Videos via Transferable Representations

Abstract

In order to learn object segmentation models in videos, conventional methods require a large amount of pixel-wise ground truth annotations. However, collecting such supervised data is time-consuming and labor-intensive. In this paper, we exploit existing annotations in source images and transfer such visual information to segment videos with unseen object categories. Without using any annotations in the target video, we propose a method to jointly mine useful segments and learn feature representations that better adapt to the target frames. The entire process is decomposed into two tasks: 1) solving a submodular function for selecting object-like segments, and 2) learning a CNN model with a transferable module for adapting seen categories in the source domain to the unseen target video. We present an iterative update scheme between two tasks to self-learn the final solution for object segmentation. Experimental results on numerous benchmark datasets show that the proposed method performs favorably against the state-of-the-art algorithms.

DAVOS: Semi-Supervised Video Object Segmentation via Adversarial Domain Adaptation

“…For that reason, the well-trained models on one dataset do not perform necessarily well when applied to another. As is stated in [21,3,30] The issue of degeneration in VOS models has existed for a while, especially for off-line methods where Figure 1: Predictions between the model trained with and without our UDA method on FBMS59 [20] and Youtube-Object [24]. The optical flow (left) indicate the motion of objects.…”