Awesome

Proposal-Free Temporal Moment Localization

Code accompanying the paper Proposal-free Temporal Moment Localization of a Natural-Language Query in Video using Guided Attention.

This repository includes:

• Code for training and testing our model for temporal moment localization in the Charades-STA and Activity-Net datasets.
• Links to the I3D features we extracted for the Charades-STA and Activity-Net which were used for the experiments in our paper.
• Links to pre-trained models on the Charades-STA and Activity-Net datasets.

Installation

1. Clone this repo

   git clone https://github.com/crodriguezo/TMLGA.git
   cd TMLGA
   

2. Create a conda environment based on our dependencies and activate it

   conda create -n <name> --file packageslist.txt
   conda activate <name>
   

   Where you can replace <name> by whatever you want.

3. Download everything

   sh ./download.sh
   

   This script will download the following things in the folder ~/data/TMLGA:

   • The glove.840B.300d.txt pretrained word embeddings.
   • The I3D features for Charades-STA and Activity-Net we extracted and used in our experiments.

   If you would like to change the default output folder for these downloads, please run sh ./download.sh <download_path>.

   This script will also install the en_core_web_md pre-trained spacy model, and download weights of our model pre-trained on the Charades-STA and Activity-Net datasets on the folders ./checkpoint/chares_sta and ./checkpoint/anet respectively.

   Downloading everything can take a while depending on your internet connection, please be patient.

Configuration

If you have modified the download path from the defaults in the script above please modify the contents of the file ./config/settings.py accordingly.

Training

To train our model in the Charades-STA dataset, please run:

python main.py --config-file=experiments/charades-sta_train.yaml

We use tensorboardX to visualize progress of our model during training. Please run the followig command to see launch tensorborad:

tensorboard --logdir=experiments/visualization/charades_sta_train/

Testing

To load our pre-trained model and test it, first make sure the weigths have been downloaded and are in the ./checkpoints/charades_sta foldel. Then simply run:

python main.py --config-file=experiments/charades-sta.yaml

Download Links

If you are interested in downloading some specific resource only, we provide the links below.

I3D Features

• Charades-STA
• Activity-Net 1 2
• YouCookII

GLoVe

• glove.840B.300d.zip

Pretrained weights

• Charades-STA

Citation

If you use our code or data please consider citing our work.

@article{opazo2019proposal,
 author = {Rodríguez-Opazo, Cristian and Marrese-Taylor, Edison and Saleh, Fatemeh Sadat and Li, Hongdong and Gould, Stephen},
 journal = {Winter Conference on Applications of Computer Vision},
 title = {Proposal-free Temporal Moment Localization of a Natural-Language Query in Video using Guided Attention},
 year = {2020}
}