Skip to content

Adaptive Short-Temporal Induced Aware Fusion Network for Driver Attention Prediction

Notifications You must be signed in to change notification settings

liuchunsense/ASIAFnet

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

10 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

ASIAF-Net


Official PyTorch implementation for paper Adaptive Short-Temporal Induced Aware Fusion Network for Predicting Attention Regions like a Driver.

ASIAF-Net is a novel driver attention prediction framework that performs self-adaptive short-temporal feature extraction. And for object-level driver attention prediction, we designed an extra object saliency estimation branch to find objects which should be focused.

Model2

Installation

This code was developed with Python 3.6 on Ubuntu 16.04. The main Python requirements:

pytorch==1.4
opencv-python
apex

Datasets

  1. Download and extract DADA-2000 Dataset.
  2. Download the Ground Truth of Object Saliency Estimation (password:enmq) and extract it.
  3. Put all the samples of the DADA-2000 datasets in the same folder, as follows
DIR
|- json_file
|    └─train_file
|    └─val_file
|    └─test_file
|- DADA_dataset
|    └─1
|      └─001
|         └─fixation
|         └─images
|         └─maps
|      |
|      └─027
|         └─fixation
|         └─images
|         └─maps
|    |
|    └─54
|      └─001
|         └─fixation
|         └─images
|         └─maps
|      |
|      └─020
|         └─fixation
|         └─images
|         └─maps
|- OSE


News 🎉

March 2022 Update - the ground truth of object salieny detection March 2022 - Release the part of code

Code

Coming Soon after publication.

About

Adaptive Short-Temporal Induced Aware Fusion Network for Driver Attention Prediction

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages