MetaTTE: a Meta-Learning Based Travel Time Estimation Model for Multi-city Scenarios

Last update: Dec 14, 2022

Related tags

Deep Learning MetaTTE

Overview

MetaTTE: a Meta-Learning Based Travel Time Estimation Model for Multi-city Scenarios

This is the official TensorFlow implementation of MetaTTE in the manuscript.

Core Requirements

tensorflow~=2.3.0
numpy~=1.18.4
spektral~=0.6.1
pandas~=1.0.3
tqdm~=4.46.0
opencv-python~=4.3.0.36
matplotlib~=3.2.1
Pillow~=7.1.2
scipy~=1.4.1

All Dependencies can be installed using the following command:

pip install -r requirements.txt

Data Preparation

We here provide the datasets we adopted in this paper with Google Drive. After downloading the zip file, please extract all the files in data directory to the data folder in this project.

Download Link: Download

Configuration

We here list a sample of our config file, and leave the comments for explanation. \ (Please DO NOT include the comments in config files)

[General]
mode = train
# Specify the absoulute path of training, validation and testing files
train_files = ./data/chengdu/train.npy,./data/porto/train.npy
val_files = ./data/chengdu/val.npy,./data/porto/val.npy
test_files = ./data/chengdu/test.npy,./data/porto/test.npy
# Specify the batch size
batch_size = 32
# Specify the number for GPU
gpu = 7
# Specify the unique label for each experiment
prefix = tte_exp_64_gru

[Model]
# Specify the inner learning rate
learning_rate = 1e-2
# Specify the inner reduce rate of learning rate
lr_reduce = 0.5
# Specify the maximum iteration
epoch = 500000
# Specify the k shot
inner_k = 10
# Specify the outer step size
outer_step_size = 0.1
# Specify the model according to the class name
model = MSMTTEGRUAttModel
# Specify the dataset according to the class name
dataset = MyDifferDatasetWithEmbedding
# Specify the dataloader according to the class name
dataloader = MyDataLoaderWithEmbedding


# mean, standard deviation for latitudes, longitudes and travel time (Chengdu is before the comma while Porto is after the comma)
[Statistics]
lat_means = 30.651168872309235,41.16060653954797
lng_means = 104.06000501543934,-8.61946359614912
lat_stds = 0.039222931811691585,0.02315827641949562
lng_stds = 0.045337940910596744,0.029208656457667292
labels_means = 1088.0075248390972,691.2889878452086
labels_stds = 1315.707363003298,347.4765869900725

Model Training

Here are commands for training the model on both Chengdu and Porto tasks.

python main.py --config=./experiments/finetuning/64/gru.conf

Eval baseline methods

Here are commands for testing the model on both Chengdu and Porto tasks.

python main.py --config=./experiments/finetuning/64/gru.conf

Citation

We currently do not provide citations.

MetaTTE: a Meta-Learning Based Travel Time Estimation Model for Multi-city Scenarios

Related tags

Overview

MetaTTE: a Meta-Learning Based Travel Time Estimation Model for Multi-city Scenarios

Core Requirements

Data Preparation

Configuration

Model Training

Eval baseline methods

Citation

Owner

morningstarwang

Dynamic Multi-scale Filters for Semantic Segmentation (DMNet ICCV'2019)

This repository gives an example on how to preprocess the data of the HECKTOR challenge

Paddle implementation for "Highly Efficient Knowledge Graph Embedding Learning with Closed-Form Orthogonal Procrustes Analysis" (NAACL 2021)

Neural Radiance Fields Using PyTorch

ImageNet-CoG is a benchmark for concept generalization. It provides a full evaluation framework for pre-trained visual representations which measure how well they generalize to unseen concepts.

Official repository for "Intriguing Properties of Vision Transformers" (2021)

IAST: Instance Adaptive Self-training for Unsupervised Domain Adaptation (ECCV 2020)

DataCLUE: 国内首个以数据为中心的AI测评（含模型分析报告）

Isaac Gym Reinforcement Learning Environments

PyQt6 configuration in yaml format providing the most simple script.

Single Image Random Dot Stereogram for Tensorflow

Real-time multi-object tracker using YOLO v5 and deep sort

Jigsaw Rate Severity of Toxic Comments

Simple torch.nn.module implementation of Alias-Free-GAN style filter and resample

Rethinking the Importance of Implementation Tricks in Multi-Agent Reinforcement Learning

This repository contains the source code and data for reproducing results of Deep Continuous Clustering paper

Multi-Scale Aligned Distillation for Low-Resolution Detection (CVPR2021)

High dimensional black-box optimizer using Latent Action Monte Carlo Tree Search algorithm

This is an official implementation for "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows" on Object Detection and Instance Segmentation.

Pytorch implementation for ACMMM2021 paper "I2V-GAN: Unpaired Infrared-to-Visible Video Translation".