CVPR 2024 "Unifying Top-down and Bottom-up Scanpath Prediction Using Transformers"

• Install Detectron2
• Install MSDeformableAttn:

  cd ./scanpath_prediction_all/sptransformer/pixel_decoder/ops
  sh make.sh
  

• Download pretrained model weights (ResNet-50 and Deformable Transformer) with the following python code

   if not os.path.exists("./pretrained_models/"):
       os.mkdir('./pretrained_models')
  
   print('downloading pretrained model weights...')
   url = f"http://vision.cs.stonybrook.edu/~cvlab_download/HAT/pretrained_models/M2F_R50_MSDeformAttnPixelDecoder.pkl"
   wget.download(url, 'pretrained_models/')
   url = f"http://vision.cs.stonybrook.edu/~cvlab_download/HAT/pretrained_models/M2F_R50.pkl"
   wget.download(url, 'pretrained_models/')
  

• Train a model with

  python train_sptransformer.py --hparams ./configs/coco_search18_dense_SSL.json --dataset-root <dataset_root> 
  

This repository contains code for scanpath prediction models for the following papers. Please cite if you use this code base.

@InProceedings{yang2024unify,
  author = {Yang, Zhibo and Mondal, Sounak and Ahn, Seoyoung and Xue, Ruoyu and Zelinsky, Gregory and Hoai, Minh and Samaras, Dimitris},
  title = {Unifying Top-down and Bottom-up Scanpath Prediction Using Transformers},
  booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  month = {June},
  year = {2024}
}