For this recipe, we simply summarize the results presented online (https://github.com/imatge-upc/activitynet-2016-cvprw/blob/master/misc/step_by_step_guide.md):
- The first step is to clone the git repository from
git clone https://github.com/imatge-upc/activitynet-2016-cvprw.git
- Then we need to download the ActivityNet v1.3 dataset, which has a size of 600 GB:
cd dataset
# This will download the videos on the default directory
sh download_videos.sh username password
# This will download the videos on the directory you specify
sh download_videos.sh username password /path/you/want
- The next step is to download the pre-trained weights for both the CNN3d and the RNN:
cd data/models
sh get_c3d_sports_weights.sh
sh get_temporal_location_weights.sh
- The last step consists of classifying the videos:
python scripts/run_all_pipeline.py -i path/to/test/video.mp4