This repository contains the implementation of the term project for 21 Fall-HKUST CSIT 5910.

We use one publicly available dataset namely CC_WEB_VIDEO , to download this dataset, you can use our developed tools in CC_WEB_VIDEO Downloader.

And we also use one self-made dataset, which is made up of videos from Bilibili and can be downloaded here.

• Clone this repo:

git clone https://github.com/Mi-Dora/PDVR_HKUST
cd PDVR_HKUST

• You can install all the dependencies by

conda install --file requirements.txt

python run_demo.py

• Triplets from the CC_WEB_VIDEO can be injected if the global features and triplet of the evaluation set are provide.

python train_dml.py --evaluation_set output_data/cc_web_video_features.npy --evaluation_triplets output_data/cc_web_video_triplets.npy --train_set output_data/vcdb_features.npy --triplets output_data/vcdb_triplets.npy --model_path model/

• Evaluate the performance of the system by providing the trained model path and the global features of the CC_WEB_VIDEO.

python evaluation.py --fusion Early --evaluation_set output_data/cc_vgg_features.npy --model_path model/

OR

python evaluation.py --fusion Late --evaluation_features cc_web_video_feature_files.txt --evaluation_set output_data/cc_vgg_features.npy --model_path model/

• The mAP and PR-curve are returned

CC_WEB_VIDEO Downloader

Near-Duplicate Video Retrieval with Deep Metric Learning

Intermediate CNN Features