Skip to content

Multiple driving action prediction model using Faster-RCNN and object-centric network.

Notifications You must be signed in to change notification settings

Xiaoyin96/SRIP19_SelfDriving

Repository files navigation

SRIP19: Self-Driving and Multi-task Learning

Content

  • BDD_action_gt: use IMU and GPS info from BDD dataset to generate single action ground truth.
  • multiple_action_labels: use AWS Mturk to label multiple actions and reasons of selected 12k BDD videos.
  • data_info: contains names of train, test and validation datasets.
  • mask-rcnn: Mask-RCNN model, forker from Facebook AI group and modified with action prediction.
  • I3D: inflated Conv3D model, adapted to Pytorch 1.0 and our new annotated BDD multi-action dataset.
  • maskrcnn-video: Using our customized I3D backbone with 640x360 image sequences input to extract glob features and roi features with selectors, performing end-to-end training.

Papers for reference

Self-Driving Review

Existing Self-Driving Datasets

Multi-task Learning methods

Video Prediction in self-driving

Some Prediction models

Summary: Video prediction papers with code

Some semantic segmentation approaches

Summary: Semantic segmentation papers with code


Useful github repo

About

Multiple driving action prediction model using Faster-RCNN and object-centric network.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 4

  •  
  •  
  •  
  •