ThreeDJCG 0.1.11

Joint Captioning and Grounding

视觉-文本实体链接使用说明

1. 环境依赖

CUDA版本: 11.7 其他依赖库的安装命令如下：

conda install pytorch==1.10.1 torchvision==0.11.2 torchaudio==0.10.1 cudatoolkit=11.3 -c pytorch -c conda-forge
pip install "git+git://github.com/erikwijmans/Pointnet2_PyTorch.git#egg=pointnet2_ops&subdirectory=pointnet2_ops_lib"

2. 下载安装

可使用如下命令下载安装算法包：

pip install -U ThreeDJCG

3. 使用示例及运行参数说明

from ThreeDJCG import ThreeDJCG

data_dict = {}
point_clouds = torch.ones([1, 10, 4]).cuda()
data_dict['point_clouds'] = point_clouds

data_dict["lang_num"] = 1
data_dict["lang_feat_list"] = torch.zeros([1, 1, 126, 300]).cuda()
data_dict["lang_len_list"] = torch.Tensor([10]).cuda()
data_dict["main_lang_feat_list"] = torch.zeros([1, 1, 126, 300]).cuda()
data_dict["main_lang_len_list"] = torch.Tensor([10]).cuda()
data_dict["first_obj_list"] = torch.Tensor([0]).cuda()
data_dict["unk_list"] = torch.zeros([1, 300]).cuda()
data_dict["unk"] = torch.zeros([1, 300]).cuda()
data_dict["istrain"] = torch.Tensor([0]).cuda()

data_dict = ThreeDJCG.inference(data_dict)