The goal of the Kinetics dataset is to help the computer vision and machine learning communities advance models for video understanding. Given this large human action classification dataset, it may be possible to learn powerful video representations that transfer to different video tasks.
: Known for intense choreography and dark plots, titles like The Man from Nowhere and I Saw the Devil are staples for genre fans.
: A meditative and beautiful exploration of life's cycles.
: Films like Silenced or Parasite use storytelling to critique class disparity and societal issues.
Whether you are a seasoned cinephile or just beginning your journey into K-movies, the depth of South Korea’s filmography ensures there is always something new and profound to discover.
: Limited screening (very rare, usually reserved for extreme content). Where to Discover and Stream
If you are looking to explore a vast catalog of hundreds of titles, several dedicated platforms offer extensive libraries:
: Known for intense choreography and dark plots, titles like The Man from Nowhere and I Saw the Devil are staples for genre fans.
: A meditative and beautiful exploration of life's cycles.
: Films like Silenced or Parasite use storytelling to critique class disparity and societal issues.
Whether you are a seasoned cinephile or just beginning your journey into K-movies, the depth of South Korea’s filmography ensures there is always something new and profound to discover.
: Limited screening (very rare, usually reserved for extreme content). Where to Discover and Stream
If you are looking to explore a vast catalog of hundreds of titles, several dedicated platforms offer extensive libraries:
1. Possible to use ImageNet checkpoints?
We allow finetuning from public ImageNet checkpoints for the supervised track -- but a link to the specific checkpoint should be provided with each submission.
2. Possible to use optical flow?
Flow can be used as long as not trained on external datasets, except if they are synthetic.
korean movies 560
3. Can we train on test data without labels (e.g. transductive)?
No.
: Known for intense choreography and dark plots,
4. Can we use semantic class label information?
Yes, for the supervised track.
several dedicated platforms offer extensive libraries:
5. Will there be special tracks for methods using fewer FLOPs / small models or just RGB vs RGB+Audio in the self-supervised track?
We will ask participants to provide the total number of model parameters and the modalities used and plan to create special mentions for those doing well in each setting, but not specific tracks.