### Install Project Dependencies (Virtual Environment) Source: https://github.com/xinshuoweng/ab3dmot/blob/master/docs/INSTALL.md Sets up a Python virtual environment named 'env', activates it, and then installs the project dependencies using pip. ```shell pip3 install venv python3 -m venv env source env/bin/activate pip3 install -r requirements.txt ``` -------------------------------- ### Install Xinshuo_PyToolbox Dependencies Source: https://github.com/xinshuoweng/ab3dmot/blob/master/docs/INSTALL.md Navigates into the cloned Xinshuo_PyToolbox directory and installs its required Python packages using pip. ```shell cd Xinshuo_PyToolbox pip3 install -r requirements.txt cd .. ``` -------------------------------- ### Install Project Dependencies (System Python) Source: https://github.com/xinshuoweng/ab3dmot/blob/master/docs/INSTALL.md Installs the required Python packages for the AB3DMOT project using pip, assuming you are using the system's Python installation. ```shell cd path/to/AB3DMOT pip3 install -r requirements.txt ``` -------------------------------- ### Clone AB3DMOT Repository Source: https://github.com/xinshuoweng/ab3dmot/blob/master/docs/INSTALL.md Clones the AB3DMOT project repository from GitHub to your local machine. ```shell git clone https://github.com/xinshuoweng/AB3DMOT.git ``` -------------------------------- ### Install Dependencies for nuScenes Source: https://github.com/xinshuoweng/ab3dmot/blob/master/docs/nuScenes.md Installs required Python packages for nuScenes experiments using pip. ```bash cd path/to/AB3DMOT source env/bin/activate cd scripts/nuScenes pip3 install -r requirements.txt cd ../.. ``` -------------------------------- ### Clone Xinshuo_PyToolbox Repository Source: https://github.com/xinshuoweng/ab3dmot/blob/master/docs/INSTALL.md Clones the Xinshuo_PyToolbox repository from GitHub, which is a dependency for AB3DMOT. ```shell git clone https://github.com/xinshuoweng/Xinshuo_PyToolbox ``` -------------------------------- ### Run AB3DMOT Demo on KITTI Source: https://github.com/xinshuoweng/ab3dmot/blob/master/README.md Executes a quick demo of the AB3DMOT method on the KITTI dataset using a pre-installed setup. This involves running the main script for detection and then post-processing scripts for thresholding and visualization. A small dataset is included, so no external download is required. ```python python3 main.py --dataset KITTI --split val --det_name pointrcnn python3 scripts/post_processing/trk_conf_threshold.py --dataset KITTI --result_sha pointrcnn_val_H1 python3 scripts/post_processing/visualization.py --result_sha pointrcnn_val_H1_thres --split val ``` -------------------------------- ### Activate Environment and Update Configuration Source: https://github.com/xinshuoweng/ab3dmot/blob/master/docs/INSTALL.md Sources the updated profile to apply the PYTHONPATH changes and then activates the Python virtual environment. ```shell source ~/.profile cd path/to/AB3DMOT source env/bin/activate ``` -------------------------------- ### Configure PYTHONPATH Environment Variable Source: https://github.com/xinshuoweng/ab3dmot/blob/master/docs/INSTALL.md Sets the PYTHONPATH environment variable to include the AB3DMOT and Xinshuo_PyToolbox directories, allowing Python to find the necessary modules. ```shell export PYTHONPATH=${PYTHONPATH}:/home/user/workspace/code/AB3DMOT export PYTHONPATH=${PYTHONPATH}:/home/user/workspace/code/AB3DMOT/Xinshuo_PyToolbox ``` -------------------------------- ### Pre-process Detections for AB3DMOT Input Source: https://github.com/xinshuoweng/ab3dmot/blob/master/docs/nuScenes.md Converts converted KITTI detection files into the input format required by the AB3DMOT main script. ```python python3 scripts/pre_processing/convert_det2input.py --dataset nuScenes --split val --det_name detname ``` -------------------------------- ### Run AB3DMOT Tracker on KITTI Test Set Source: https://github.com/xinshuoweng/ab3dmot/blob/master/docs/KITTI.md Executes the AB3DMOT tracker on the KITTI MOT test set using PointRCNN detections. The results are saved in a specified output folder. ```bash python3 main.py --dataset KITTI --det_name pointrcnn --split test ``` -------------------------------- ### Visualize 3D MOT Qualitative Results Source: https://github.com/xinshuoweng/ab3dmot/blob/master/docs/KITTI.md Generates visualizations of the 3D MOT system's performance on images from the KITTI dataset. This process requires OpenCV 3 and saves visual outputs to specified directories. ```python python3 scripts/post_processing/visualization.py --dataset KITTI --result_sha pointrcnn_test_H1_thres --split test ``` -------------------------------- ### Run AB3DMOT Inference on nuScenes Source: https://github.com/xinshuoweng/ab3dmot/blob/master/docs/nuScenes.md Executes the AB3DMOT tracker on the nuScenes dataset using pre-computed detections. ```python python3 main.py --dataset nuScenes --det_name megvii --split val python3 main.py --dataset nuScenes --det_name centerpoint --split val ``` -------------------------------- ### Evaluate 3D MOT on nuScenes Test Set Source: https://github.com/xinshuoweng/ab3dmot/blob/master/docs/nuScenes.md Runs the main script to evaluate 3D MOT performance on the nuScenes test set. The results are then compressed and uploaded to the official nuScenes evaluation server for final assessment. ```shell python3 main.py --dataset nuScenes --det_name megvii --split test python3 scripts/nuScenes/export_kitti.py kitti_trk_result2nuscenes --result_name megvii_test_H1 --split test ``` -------------------------------- ### Reproduce 2D MOT Evaluation on KITTI Validation Set Source: https://github.com/xinshuoweng/ab3dmot/blob/master/docs/KITTI.md Executes the evaluation script for 2D MOT on the KITTI validation set using the PointRCNN detector. The output metrics should match the provided table, with FPS potentially varying based on the hardware. ```python python3 scripts/KITTI/evaluate.py pointrcnn_val_H1 1 2D 0.5 ``` -------------------------------- ### Run AB3DMOT Tracker on KITTI Validation Set Source: https://github.com/xinshuoweng/ab3dmot/blob/master/docs/KITTI.md Executes the AB3DMOT tracker on the KITTI MOT validation set using PointRCNN detections. This command generates tracking results for all categories and individual categories. ```bash python3 main.py --dataset KITTI --det_name pointrcnn ``` -------------------------------- ### Compress and Upload KITTI Test Set Results Source: https://github.com/xinshuoweng/ab3dmot/blob/master/docs/KITTI.md Compresses the generated tracking results for the KITTI test set, which is necessary for submission to the official KITTI 2D MOT evaluation server. ```bash ./results/KITTI/pointrcnn_test_H1_thres/data_0 ``` -------------------------------- ### Reproduce 2D MOT Evaluation on KITTI Test Set Source: https://github.com/xinshuoweng/ab3dmot/blob/master/docs/KITTI.md Runs a post-processing script to generate results for 2D MOT evaluation on the KITTI test set. The output requires compression and upload to the KITTI evaluation server for official scoring, which may not include custom metrics. ```python python3 scripts/post_processing/trk_conf_threshold.py --dataset KITTI --result_sha pointrcnn_test_H1 ``` -------------------------------- ### Quick 3D MOT Evaluation on nuScenes Validation Set (Unofficial) Source: https://github.com/xinshuoweng/ab3dmot/blob/master/docs/nuScenes.md Provides a faster, unofficial evaluation script for the nuScenes validation set, adapted from KITTI evaluation. Note that results from this script may differ from the official evaluation due to implementation variations. ```shell python3 scripts/nuScenes/evaluate_quick.py megvii_val_H1 1 val ``` -------------------------------- ### Convert nuScenes Detections to KITTI Format Source: https://github.com/xinshuoweng/ab3dmot/blob/master/docs/nuScenes.md Converts nuScenes object detection results (in nuScenes format) to the KITTI detection format. ```python python3 scripts/nuScenes/export_kitti.py nuscenes_obj_result2kitti --result_name detname --split val ``` -------------------------------- ### Evaluate AB3DMOT 3D MOT Results on KITTI Validation Set Source: https://github.com/xinshuoweng/ab3dmot/blob/master/docs/KITTI.md Evaluates the AB3DMOT 3D Multi-Object Tracking results on the KITTI MOT validation set using different 3D IoU thresholds (0.25, 0.5, 0.7). This script reproduces the quantitative results reported in the paper. ```bash python3 scripts/KITTI/evaluate.py pointrcnn_val_H1 1 3D 0.25 ``` ```bash python3 scripts/KITTI/evaluate.py pointrcnn_val_H1 1 3D 0.5 ``` ```bash python3 scripts/KITTI/evaluate.py pointrcnn_Car_val_H1 1 3D 0.7 ``` -------------------------------- ### Convert nuScenes GT to KITTI Format Source: https://github.com/xinshuoweng/ab3dmot/blob/master/docs/nuScenes.md Converts nuScenes ground truth data into the KITTI tracking format for processing. ```python python3 scripts/nuScenes/export_kitti.py nuscenes_gt2kitti_trk --split val python3 scripts/nuScenes/export_kitti.py nuscenes_gt2kitti_trk --split test ``` -------------------------------- ### Convert Results and Evaluate 3D MOT on nuScenes Validation Set Source: https://github.com/xinshuoweng/ab3dmot/blob/master/docs/nuScenes.md Converts detection results to the nuScenes tracking format and runs the official nuScenes evaluation code for the validation set. This process aims to reproduce quantitative 3D MOT results. ```shell python3 scripts/nuScenes/export_kitti.py kitti_trk_result2nuscenes --result_name megvii_val_H1 --split val python3 scripts/nuScenes/evaluate.py --result_path ./results/nuScenes/megvii_val_H1/results_val.json ``` -------------------------------- ### Filter Tracklets and Visualize Results (Python) Source: https://github.com/xinshuoweng/ab3dmot/blob/master/docs/nuScenes.md These scripts are used to process and visualize the results of a 3D MOT system on the nuScenes dataset. The first script filters out low-score tracklets, and the second generates image and video visualizations. OpenCV 3 is a required dependency. ```python python3 scripts/post_processing/trk_conf_threshold.py --dataset nuScenes-- result_sha megvii_val_H1 ``` ```python python3 scripts/post_processing/visualization.py --dataset nuScenes --result_sha megvii_val_H1_thres --split val ``` === COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.