Traffic4D: Single View Reconstruction of Repetitious Activity Using Longitudinal Self-Supervision

Related tags

Deep LearningTraffic4D-Release
Overview

Traffic4D: Single View Reconstruction of Repetitious Activity Using Longitudinal Self-Supervision

Project | PDF | Poster
Fangyu Li, N. Dinesh Reddy, Xudong Chen and Srinivasa G. Narasimhan
Proceedings of IEEE Intelligent Vehicles Symposium (IV'21)
Best Paper Award

Following instructions below, the user will get keypoints, trajectory reconstruction and vehicular activity clustering results like

Set up

The set up process can be skipped if using docker. Please check "Docker" section.

Python

Python version 3.6.9 is used. Python packages are in requirements.txt .

git clone https://github.com/Emrys-Lee/Traffic4D-Release.git
sudo apt-get install python3.6
sudo apt-get install python3-pip
cd Traffic4D-Release
pip3 install -r requirements.txt

C++

Traffic4D uses C++ libraries ceres and pybind for efficient optimization. pybind needs clang compiler, so Traffic4D uses clang compiler.

Install clang compiler

sudo apt-get install clang++-6.0

Install prerequisites for ceres

# CMake
sudo apt-get install cmake
# google-glog + gflags
sudo apt-get install libgoogle-glog-dev libgflags-dev
# BLAS & LAPACK
sudo apt-get install libatlas-base-dev
# Eigen3
sudo apt-get install libeigen3-dev
# SuiteSparse and CXSparse (optional)
sudo apt-get install libsuitesparse-dev

Download and install ceres

wget https://github.com/ceres-solver/ceres-solver/archive/1.12.0.zip
unzip 1.12.0.zip
cd ceres-solver-1.12.0/
mkdir build
cd build
cmake ..
make
sudo make install

Download and install pybind

git clone https://github.com/pybind/pybind11
cd pybind11
cmake .
make
sudo make install

Build Traffic4D optimization library

cd Traffic4D-Release/src/ceres
make

ceres_reconstruct.so and ceres_spline.so are generated under path Traffic4D-Release/src/ceres/.

Dataset

Download dataset and pre-generated results from here, and put it under Traffic4D-Release/.

cd Traffic4D-Release
mv Data-Traffic4D.zip ./
unzip Data-Traffic4D.zip

The directory should be like

Traffic4D-Release/
    Data-Traffic4D/
    └───fifth_morewood/
        └───fifth_morewood_init.vd
        └───top_view.png
        └───images/
                00001.jpg
                00002.jpg
                ...
                06288.jpg
    └───arterial_kennedy/
        └───arterial_kennedy_init.vd
        └───top_view.png
        └───images/
                <put AI City Challenge frames here>
        ...

The input and output paths can be modified in config/*.yml.

Explanation

1. Input videos

Sample videos in Traffic4D are provided. Note arterial_kennedy and dodge_century are from Nvidia AI City Challenge City-Scale Multi-Camera Vehicle Tracking Challenge Track. Please request the access to the dataset here. Once get the data, run

ffmpeg -i <mtmc-dir>/train/S01/c001/vdo.avi Traffic4D-Release/Data-Traffic4D/arterial_kennedy/images/%05d.jpg
ffmpeg -i <mtmc-dir>/test/S02/c007/vdo.avi Traffic4D-Release/Data-Traffic4D/dodge_century/images/%05d.jpg

to extract frames into images/.

2. Pre-Generated 2D results

Detected 2D bounding boxes, keypoints and tracking IDs are stored in *_init.vd. Check Occlusionnet implementation for detecting keypoints; V-IOU for multi-object tracking.

3. Output folder

Folder Traffic4D-Release/Result/ will be created by default.

Experiments

Run python exp/traffic4d.py config/<intersection_name>.yml <action>. Here YML configuration files for multiple intersections are provided under config/ folder. <action> shoulbe be reconstruction or clustering to perform longitudinal reconstruction and activity clustering sequentially. For example, below runs Fifth and Morewood intersection.

cd Traffic4D-Release
python3 exp/traffic4d.py config/fifth_morewood.yml reconstruction
python3 exp/traffic4d.py config/fifth_morewood.yml clustering

Results

Find these results in the output folder:

  1. 2D keypoints: If 3D reconstruction is done, 2D reprojected keypoints will be plotted in Traffic4D-Release/Result/<intersection_name>_keypoints/.
  2. 3D reconstructed trajectories and clusters: The clustered 3D trajectories are plotted on the top view map as Traffic4D-Release/Result/<intersection_name>_top_view.jpg.

Docker

We provide docker image with dependencies already set up. The steps in "Set up" can be skipped if you use docker image. You still need to clone the repo and download the dataset and put it in under Traffic4D-Release/.

git clone https://github.com/Emrys-Lee/Traffic4D-Release.git

Pull Traffic4D docker image.

docker pull emrysli/traffic4d-release:latest

Then create a container and map the git repo into docker container to access the dataset. For example, if the cloned repo locates at host directory /home/xxx/Traffic4D-Release, <path_to_repo> should be /home/xxx. If <path_in_container> is /home/yyy, then /home/xxx/Traffic4D-Release will be mapped as /home/yyy/Traffic4D-Release inside the container.

docker run -it -v <path_to_repo>/Traffic4D-Release:<path_in_container>/Traffic4D-Release emrysli/traffic4d-release:latest /bin/bash

Inside container compile Traffic4D again.

# inside container
cd <path_in_container>/Traffic4D-Release/src/ceres
make

Run experiments.

cd <path_in_container>/Traffic4D-Release
python3 exp/traffic4d.py config/fifth_morewood.yml reconstruction
python3 exp/traffic4d.py config/fifth_morewood.yml clustering

Trouble Shooting

  1. tkinter module is missing
File "/usr/local/lib/python3.6/dist-packages/matplotlib/backends/_backend_tk.py", line 5, in <module>
    import tkinter as Tk
ModuleNotFoundError: No module named 'tkinter'

Solution: install tkinter.

sudo apt-get install python3-tk
  1. opencv import error such as
File "/usr/local/lib/python3.6/dist-packages/cv2/__init__.py", line 3, in <module>
    from .cv2 import *
ImportError: libSM.so.6: cannot open shared object file: No such file or directory

Solution: install the missing libraries.

sudo apt-get install libsm6 libxrender1 libfontconfig1 libxext6

Citation

Traffic4D

@conference{Li-2021-127410,
author = {Fangyu Li and N. Dinesh Reddy and Xudong Chen and Srinivasa G. Narasimhan},
title = {Traffic4D: Single View Reconstruction of Repetitious Activity Using Longitudinal Self-Supervision},
booktitle = {Proceedings of IEEE Intelligent Vehicles Symposium (IV '21)},
year = {2021},
month = {July},
publisher = {IEEE},
keywords = {Self-Supervision, vehicle Detection, 4D Reconstruction, 3D reconstuction, Pose Estimation.},
}

Occlusion-Net

@inproceedings{onet_cvpr19,
author = {Reddy, N. Dinesh and Vo, Minh and Narasimhan, Srinivasa G.},
title = {Occlusion-Net: 2D/3D Occluded Keypoint Localization Using Graph Networks},
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
pages = {7326--7335},
year = {2019}
}
Owner
GitHub Repository
Code for our method RePRI for Few-Shot Segmentation. Paper at http://arxiv.org/abs/2012.06166

Region Proportion Regularized Inference (RePRI) for Few-Shot Segmentation In this repo, we provide the code for our paper : "Few-Shot Segmentation Wit

Malik Boudiaf 138 Dec 12, 2022
A deep learning framework for historical document image analysis

DIVA-DAF Description A deep learning framework for historical document image analysis. How to run Install dependencies # clone project git clone https

9 Aug 04, 2022
Pytorch implementation of COIN, a framework for compression with implicit neural representations 🌸

COIN 🌟 This repo contains a Pytorch implementation of COIN: COmpression with Implicit Neural representations, including code to reproduce all experim

Emilien Dupont 104 Dec 14, 2022
Official Python implementation of the FuzionCoin protocol

PyFuzc Official Python implementation of the FuzionCoin protocol WARNING: Under construction. Use at your own risk. Some functions may not work. Setup

FuzionCoin 3 Jul 07, 2022
Modifications of the official PyTorch implementation of StyleGAN3. Let's easily generate images and videos with StyleGAN2/2-ADA/3!

Alias-Free Generative Adversarial Networks (StyleGAN3) Official PyTorch implementation of the NeurIPS 2021 paper Alias-Free Generative Adversarial Net

Diego Porres 185 Dec 24, 2022
Semi-Supervised Learning with Ladder Networks in Keras. Get 98% test accuracy on MNIST with just 100 labeled examples !

Semi-Supervised Learning with Ladder Networks in Keras This is an implementation of Ladder Network in Keras. Ladder network is a model for semi-superv

Divam Gupta 101 Sep 07, 2022
PyTorch implementation of Neural Dual Contouring.

NDC PyTorch implementation of Neural Dual Contouring. Citation We are still writing the paper while adding more improvements and applications. If you

Zhiqin Chen 140 Dec 26, 2022
Sequence-to-Sequence learning using PyTorch

Seq2Seq in PyTorch This is a complete suite for training sequence-to-sequence models in PyTorch. It consists of several models and code to both train

Elad Hoffer 514 Nov 17, 2022
Neural Style and MSG-Net

PyTorch-Style-Transfer This repo provides PyTorch Implementation of MSG-Net (ours) and Neural Style (Gatys et al. CVPR 2016), which has been included

Hang Zhang 904 Dec 21, 2022
KakaoBrain KoGPT (Korean Generative Pre-trained Transformer)

KoGPT KoGPT (Korean Generative Pre-trained Transformer) https://github.com/kakaobrain/kogpt https://huggingface.co/kakaobrain/kogpt Model Descriptions

Kakao Brain 799 Dec 28, 2022
Tensorflow implementation of Character-Aware Neural Language Models.

Character-Aware Neural Language Models Tensorflow implementation of Character-Aware Neural Language Models. The original code of author can be found h

Taehoon Kim 751 Dec 26, 2022
Optimized primitives for collective multi-GPU communication

NCCL Optimized primitives for inter-GPU communication. Introduction NCCL (pronounced "Nickel") is a stand-alone library of standard communication rout

NVIDIA Corporation 2k Jan 09, 2023
Code for STFT Transformer used in BirdCLEF 2021 competition.

STFT_Transformer Code for STFT Transformer used in BirdCLEF 2021 competition. The STFT Transformer is a new way to use Transformers similar to Vision

Jean-François Puget 69 Sep 29, 2022
Weakly Supervised 3D Object Detection from Point Cloud with Only Image Level Annotation

SCCKTIM Weakly Supervised 3D Object Detection from Point Cloud with Only Image-Level Annotation Our code will be available soon. The class knowledge t

1 Nov 12, 2021
Mesh TensorFlow: Model Parallelism Made Easier

Mesh TensorFlow - Model Parallelism Made Easier Introduction Mesh TensorFlow (mtf) is a language for distributed deep learning, capable of specifying

1.3k Dec 26, 2022
This project aims to segment 4 common retinal lesions from Fundus Images.

This project aims to segment 4 common retinal lesions from Fundus Images.

Husam Nujaim 1 Oct 10, 2021
PyTorch implementation of the supervised learning experiments from the paper Model-Agnostic Meta-Learning (MAML)

pytorch-maml This is a PyTorch implementation of the supervised learning experiments from the paper Model-Agnostic Meta-Learning (MAML): https://arxiv

Kate Rakelly 516 Jan 05, 2023
Official implementation for ICDAR 2021 paper "Handwritten Mathematical Expression Recognition with Bidirectionally Trained Transformer"

Handwritten Mathematical Expression Recognition with Bidirectionally Trained Transformer Description Convert offline handwritten mathematical expressi

Wenqi Zhao 87 Dec 27, 2022
A simple command line tool for text to image generation, using OpenAI's CLIP and a BigGAN.

Ryan Murdock has done it again, combining OpenAI's CLIP and the generator from a BigGAN! This repository wraps up his work so it is easily accessible to anyone who owns a GPU.

Phil Wang 2.3k Jan 09, 2023
PyTorch implementation of SIFT descriptor

This is an differentiable pytorch implementation of SIFT patch descriptor. It is very slow for describing one patch, but quite fast for batch. It can

Dmytro Mishkin 150 Dec 24, 2022