Bin Yan

@kongbia Hi, thank you for your interest in our work. We will provide all raw results and complete models after this paper is accepted.

@maojiaoli 你好，在跟踪领域如果想做分割（预测mask）的话，可以使用一些带有mask标注的数据集，例如COCO Instance Segmentation， Youtube-VOS， 以及一些“显著性检测”的数据集（DUTS，DUT-OMRON等）

Hi, please post the detailed error log or pictures.

Hi, as mentioned in the README, the environment is: NVIDIA RTX-2080Ti, Ubuntu 16.04, CUDA 10.0 We don't suggest running Deep Learning on a virtual machine. You can try installing a...

Please check whether your computer has Nvidia GPUs and whether you have installed the corresponding graphic card driver.

Hi, do you mean the interaction module [here](https://github.com/MasterBin-IIAU/Unicorn/blob/e3653997cc56848df67084b3134806feb7191c57/unicorn/exp/unicorn_track.py#L132)? The interaction module is applied to enhance embeddings for more discriminative correspondence as mentioned in section 3.2

@xinsuinizhuan sorry, for now, the codebase does not support the functions you mentioned. Unicorn uses some operations (such as Depth-wise convolution, Group Normalization, and Deformable Attention), which may be unfriendly...

@qyu21490 Hi, we will release a demo script as soon as possible. This script will show how to use Unicorn for inference of four tracking tasks.

Hi, it is possible to use your own detector for detecting objects and then use our embeddings for the association. The embedding is obtained by extracting features from the centers...

@972821054 Hi, as mentioned in the [model_zoo.md](https://github.com/MasterBin-IIAU/Unicorn/blob/master/assets/model_zoo.md), please download pre-trained weights and put it under the Unicorn_outputs folder.