A Simple PyTorch Implementation of the Semantic Completion Networks

The paper “Weakly-Supervised Video Moment Retrieval via Semantic Completion Network” accepted by AAAI-2020 is available here pdf.

• Train on the ActivityNet Dataset:

python train.py --config-path config/activitynet/main.json

Prerequisites

• pytorch
• python3
• gensim
• opencv-python
• h5py

A link to pre-extracted features

The C3D visual features from google drive.