Kernelized Memory Network for Video Object Segmentation

ECCV 2020  ·  Hongje Seong, Junhyuk Hyun, Euntai Kim ·

Semi-supervised video object segmentation (VOS) is a task that involves predicting a target object in a video when the ground truth segmentation mask of the target object is given in the first frame. Recently, space-time memory networks (STM) have received significant attention as a promising solution for semi-supervised VOS. However, an important point is overlooked when applying STM to VOS. The solution (STM) is non-local, but the problem (VOS) is predominantly local. To solve the mismatch between STM and VOS, we propose a kernelized memory network (KMN). Before being trained on real videos, our KMN is pre-trained on static images, as in previous works. Unlike in previous works, we use the Hide-and-Seek strategy in pre-training to obtain the best possible results in handling occlusions and segment boundary extraction. The proposed KMN surpasses the state-of-the-art on standard benchmarks by a significant margin (+5% on DAVIS 2017 test-dev set). In addition, the runtime of KMN is 0.12 seconds per frame on the DAVIS 2016 validation set, and the KMN rarely requires extra computation, when compared with STM.

PDF Abstract ECCV 2020 PDF ECCV 2020 Abstract
Task Dataset Model Metric Name Metric Value Global Rank Uses Extra
Training Data
Result Benchmark
Semi-Supervised Video Object Segmentation DAVIS 2016 KMN Jaccard (Mean) 89.5 # 28
F-measure (Mean) 91.5 # 28
J&F 90.5 # 28
Semi-Supervised Video Object Segmentation DAVIS 2017 (test-dev) KMN J&F 77.2 # 28
Jaccard (Mean) 74.1 # 27
F-measure (Mean) 80.3 # 28
Semi-Supervised Video Object Segmentation DAVIS 2017 (val) KMN Jaccard (Mean) 80 # 34
F-measure (Mean) 85.6 # 35
J&F 82.8 # 34
Semi-Supervised Video Object Segmentation DAVIS (no YouTube-VOS training) KMN FPS 8.33 # 16
D16 val (G) 87.6 # 3
D16 val (J) 87.1 # 5
D16 val (F) 88.1 # 3
D17 val (G) 76.0 # 7
D17 val (J) 74.2 # 7
D17 val (F) 77.8 # 8
Semi-Supervised Video Object Segmentation YouTube-VOS 2018 KMN F-Measure (Seen) 85.6 # 38
F-Measure (Unseen) 83.3 # 38
Overall 81.4 # 37
Jaccard (Seen) 81.4 # 35
Jaccard (Unseen) 75.3 # 37

Methods