Command Palette
Search for a command to run...
{Arnold W. M. Smeulders Efstratios Gavves Zhenyang Li Ran Tao Cees G. M. Snoek}
Abstract
This paper strives to track a target object in a video. Rather than specifying the target in the first frame of a video by a bounding box, we propose to track the object based on a natural language specification of the target, which provides a more natural human-machine interaction as well as a means to improve tracking results. We define three variants of tracking by language specification: one relying on lingual target specification only, one relying on visual target specification based on language, and one leveraging their joint capacity. To show the potential of tracking by natural language specification we extend two popular tracking datasets with lingual descriptions and report experiments. Finally, we also sketch new tracking scenarios in surveillance and other live video streams that become feasible with a lingual specification of the target.
Benchmarks
| Benchmark | Methodology | Metrics |
|---|---|---|
| referring-expression-segmentation-on-a2d | Li et al. | AP: 0.163 IoU mean: 0.354 IoU overall: 0.515 Precision@0.5: 0.387 Precision@0.6: 0.290 Precision@0.7: 0.175 Precision@0.8: 0.066 Precision@0.9: 0.001 |
| referring-expression-segmentation-on-j-hmdb | Li et al. | AP: 0.173 IoU mean: 0.491 IoU overall: 0.529 Precision@0.5: 0.578 Precision@0.6: 0.335 Precision@0.7: 0.103 Precision@0.8: 0.060 Precision@0.9: 0.000 |
Build AI with AI
From idea to launch — accelerate your AI development with free AI co-coding, out-of-the-box environment and best price of GPUs.