The 2017 DAVIS Challenge on Video Object Segmentation is a public dataset, benchmark, and competition designed to advance the field of video object segmentation. Building on the success of previous initiatives like ILSVRC and PASCAL VOC, this challenge aims to foster research and innovation in video object segmentation. The challenge includes a new, larger dataset (DAVIS 2017) with 150 sequences and 10,459 annotated frames, featuring more complex scenes with multiple objects, smaller objects, fine structures, occlusions, and fast motion. The evaluation metrics used are region similarity ($\mathcal{J}$) and contour accuracy ($\mathcal{F}$), with the overall performance measured as the average of these metrics over all object instances. The challenge received entries from 22 teams, leading to a 20% improvement in the state-of-the-art performance. The paper provides a detailed analysis of the results, highlighting the strengths and weaknesses of the participating methods, and discusses the challenges posed by multiple objects and small objects in video object segmentation.The 2017 DAVIS Challenge on Video Object Segmentation is a public dataset, benchmark, and competition designed to advance the field of video object segmentation. Building on the success of previous initiatives like ILSVRC and PASCAL VOC, this challenge aims to foster research and innovation in video object segmentation. The challenge includes a new, larger dataset (DAVIS 2017) with 150 sequences and 10,459 annotated frames, featuring more complex scenes with multiple objects, smaller objects, fine structures, occlusions, and fast motion. The evaluation metrics used are region similarity ($\mathcal{J}$) and contour accuracy ($\mathcal{F}$), with the overall performance measured as the average of these metrics over all object instances. The challenge received entries from 22 teams, leading to a 20% improvement in the state-of-the-art performance. The paper provides a detailed analysis of the results, highlighting the strengths and weaknesses of the participating methods, and discusses the challenges posed by multiple objects and small objects in video object segmentation.