Testing script and data needed to run it for NeurIPS 2024 submission 18634

Dependencies: pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121 (get specific command for your environment from Pytorch website. If you can't use CUDA, you must edit the code to use the CPU.) pip install timm pip install opencv-python pip install pillow pip install joblib pip install scipy pip install scikit-learn Please make the following edits to run the script. Also, either move the models folder (https://zenodo.org/records/11245477) or change the directory. Also, add a folder to the directory with the testing script called "Results" The testing script also includes the models for directionality, another statistic we attempted to predict. We did not report the values for directionality in our paper, so feel free to comment lines 2175-2253 and lines 2296-2308 to speed up the computation for the statistics presented in the paper. Due to time constraints, this version also doesn't include the class-modifying arguments for Brownian and straight-trajectory motion. These will be included in the testing script on the official release of DeepTrakStat on Github. line 485: sim_directory='ground_truth_trajectories/' line 486: tmate_directory='tmate_trajectories/' lines 509-525: directories = [ 'simulated_imagery/1000part_16xspeed_heterogeneous/','simulated_imagery/1000part_32xspeed_heterogeneous/', 'simulated_imagery/1000part_4xspeed_heterogeneous/', 'simulated_imagery/2000part_16xspeed_heterogeneous/','simulated_imagery/2000part_32xspeed_heterogeneous/', ## Group 1 'simulated_imagery/500part_16xspeed_heterogeneous/','simulated_imagery/500part_32xspeed_heterogeneous/', 'simulated_imagery/500part_4xspeed_heterogeneous/', 'simulated_imagery/sim111_brown/', 'simulated_imagery/sim112_brown/', 'simulated_imagery/sim113_brown/','simulated_imagery/sim114_brown/', # Group 2 'simulated_imagery/sim115_brown/','simulated_imagery/sim119_brown/','simulated_imagery/sim120_brown/', 'simulated_imagery/sim2201/', 'simulated_imagery/sim2210/', 'simulated_imagery/sim2215/', 'simulated_imagery/sim2220/','simulated_imagery/sim2230/','simulated_imagery/sim2235/','simulated_imagery/sim2240/', # Group 3 'simulated_imagery/sim2241/','simulated_imagery/sim2242/','simulated_imagery/sim2243/','simulated_imagery/sim2244/', #Group 4 'simulated_imagery/sim2251/','simulated_imagery/sim2252/','simulated_imagery/sim2253/', 'simulated_imagery/sim2254/','simulated_imagery/sim2255/','simulated_imagery/sim2256/', 'simulated_imagery/sim2257/','simulated_imagery/sim2258/','simulated_imagery/sim2259/','simulated_imagery/sim2260/', # Group 5 'simulated_imagery/test1/','simulated_imagery/test2/','simulated_imagery/test5/','simulated_imagery/test8/' #Group 6 ] line 606: sorted_frames = sorted(files, key=lambda x: int(x[4:-4]))

### 依赖配置
执行以下命令安装核心依赖：`pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121`（请从PyTorch官方网站获取适配您运行环境的具体安装命令；若无法使用CUDA加速，请修改代码以适配CPU运行）。
此外请依次执行以下命令安装其余依赖：`pip install timm`、`pip install opencv-python`、`pip install pillow`、`pip install joblib`、`pip install scipy`、`pip install scikit-learn`。

### 运行前置配置
运行本脚本前请完成以下两项调整：
1. 请将模型文件夹（可从https://zenodo.org/records/11245477 获取）移动至指定目录，或修改代码中的目录路径；
2. 请在存放测试脚本的目录下新建名为`Results`的文件夹。

### 计算加速说明
本测试脚本同时包含了方向性（我们尝试预测的另一项统计指标）的相关模型。由于论文中未报告方向性的相关数值，您可自由注释第2175至2253行以及第2296至2308行代码，以加速论文中提及的统计指标的计算流程。

### 版本说明
受限于开发时间，当前版本未包含布朗运动与直线轨迹运动的类别修改参数，该部分内容将在DeepTrakStat的GitHub官方发布版本的测试脚本中补充完整。

### 代码行修改指引
1. 第485行：`sim_directory='ground_truth_trajectories/'`
2. 第486行：`tmate_directory='tmate_trajectories/'`
3. 第509至525行：
`directories = [ 'simulated_imagery/1000part_16xspeed_heterogeneous/','simulated_imagery/1000part_32xspeed_heterogeneous/', 'simulated_imagery/1000part_4xspeed_heterogeneous/', 'simulated_imagery/2000part_16xspeed_heterogeneous/','simulated_imagery/2000part_32xspeed_heterogeneous/', ## 组1 'simulated_imagery/500part_16xspeed_heterogeneous/','simulated_imagery/500part_32xspeed_heterogeneous/', 'simulated_imagery/500part_4xspeed_heterogeneous/', 'simulated_imagery/sim111_brown/', 'simulated_imagery/sim112_brown/', 'simulated_imagery/sim113_brown/','simulated_imagery/sim114_brown/', # 组2 'simulated_imagery/sim115_brown/','simulated_imagery/sim119_brown/','simulated_imagery/sim120_brown/', 'simulated_imagery/sim2201/', 'simulated_imagery/sim2210/', 'simulated_imagery/sim2215/', 'simulated_imagery/sim2220/','simulated_imagery/sim2230/','simulated_imagery/sim2235/','simulated_imagery/sim2240/', # 组3 'simulated_imagery/sim2241/','simulated_imagery/sim2242/','simulated_imagery/sim2243/','simulated_imagery/sim2244/', #组4 'simulated_imagery/sim2251/','simulated_imagery/sim2252/','simulated_imagery/sim2253/', 'simulated_imagery/sim2254/','simulated_imagery/sim2255/','simulated_imagery/sim2256/', 'simulated_imagery/sim2257/','simulated_imagery/sim2258/','simulated_imagery/sim2259/','simulated_imagery/sim2260/', # 组5 'simulated_imagery/test1/','simulated_imagery/test2/','simulated_imagery/test5/','simulated_imagery/test8/' #组6 ]`
4. 第606行：`sorted_frames = sorted(files, key=lambda x: int(x[4:-4]))`