The GOT-10k dataset contains more than 10,000 video segments of real-world moving objects and over 1.5 million manually labeled bounding boxes. The test set embodies 84 object classes and 32 motion classes with only 180 video segments, allowing for efficient evaluation. The dataset provides extra labels including object visible ratios and motion classes as additional supervision for handling specific challenges.
For more information please read GOT-10k: A Large High-Diversity Benchmark for Generic Object Tracking in the Wild.
Each sequence folder contains 4 annotation files and 1 meta file. A brief description of these files follows (let N denotes sequence length):
Values 0~8 in file cover.label correspond to ranges of object visible ratios: 0%, (0%, 15%], (15%~30%], (30%, 45%], (45%, 60%], (60%, 75%], (75%, 90%], (90%, 100%) and 100% respectively.
The dataset can be downloaded here. Those interested need to fill a download form with their institutional address and they shall receive the data.
The authors offer a selection of sample Python and Matlab code in their respective github repositories.
Models provided by community contributors are presented in the dataset benchmark. Note that not all models have a published architecture (in fact, most of them do not have any additional information besides their name).
The official benchmark can be consulted here.
No associated challenges have been found. However, those interested in the dataset can contribute with their own models.
Dataset ditributed under the CC-BY-NC-SA 4.0 license.