14 datasets found

The Davis dataset contains fifty high-resolution videos with pixel-accurate ground truth.

The dataset used for video segmentation, including Moving MNIST, Change detection, Segtrack V2, and Davis datasets.

A panoramic video object segmentation dataset, providing 150 videos with high video resolutions and diverse motions.

LVOS is a long-term video object segmentation benchmark dataset, consisting of 220 videos with a total duration of 421 minutes.

Video segmentation requires consistently segmenting and tracking objects over time. Due to the quadratic dependency on input size, directly applying self-attention to video...

The INRIA YouTube Instructional Videos dataset contains five tasks of different instructional domains: “making coffee”, “changing a car tire”, “CPR”, “jumping a car”, and...

The Breakfast Actions dataset contains 70 hours of cooking activities of varying complexity. It contains 10 different cooking tasks (with about 170 videos per task), which can...

Video object segmentation is challenging due to the factors like rapidly fast motion, cluttered backgrounds, arbitrary object appearance variation and shape deformation.

The DAVIS-17 dataset, which is a benchmark for video object segmentation.

The dataset used for training and testing the OMG-Seg model, which includes COCO panoptic, COCO-SAM, VIPSeg, Youtube-VIS-2019, and Youtube-VIS-2021 datasets.

The VIPSeg dataset contains videos with annotations for video panoptic segmentation.

The dataset used in the paper is a pre-trained ResNet-50 classifier, which is used for image synthesis, unpaired image-to-image translation, and feature similarity estimation.

The COIN dataset is a large-scale instructional video dataset that contains 100 hours of video. The dataset is used for instructional video analysis and understanding.

The Youtube-VOS dataset is a large-scale benchmark for multi-object video object segmentation, containing 120 videos with dense annotation.