Deep Learning Camp Jeju is an annual camp held at Jeju Island, South Korea, which was first launched in 2017 with 20 participants from all over the world.

This year, it will again be hosting 20-30 participants, that come from various backgrounds but all interested in practicing and advancing deep learning.

Why should I apply?

How does this work?

Why Jeju?

Important Dates

Date (UTC) Description
2018-04-02 (Mon) Applications open
2018-04-30 (Mon) Applications closed
2018-05-23 (Wed) Teams announced (email was sent)
2018-06-01 (Fri) Start camp online
2018-07-01 (Sun) Gather at Jeju
2018-07-02 (Mon) Start camp offline
2018-07-13 (Fri) Mid-term conference & presentation
2018-07-27 (Fri) Final presentation day!



How to apply

We are accepting applications through the following form:

[Click me!]

Applications will be accepted until April 30th, 2018 (UTC) (or precisely 2018-04-30T23:59:59Z).

In your timezone, it's


  1. Can I propose more than one project?
    • Sure! Just include both topics in your application.
  2. Can I apply as a team?
    • We value teamwork but are currently seeking for individual participants.
  3. Can I join the camp with my own money if my proposal is not selected?
    • No, because we have limited space, but you may consider applying to the Deep learning summer school and/or hackathon which will have separate application processes.
  4. Will datasets be provided?
    • No, the participant should prepare their own datasets.
  5. Can we just implement previously existing models, or should we have new ideas?
    • It’s okay to just do an implementation of a model, but it is definitely preferred to have an idea to improve that model or build a real-life application using it.
  6. Am I allowed to bring a partner?
    • Unfortunately, we only provide accommodations for the participants. You can book your own accommodations if you wish to stay with your partner during the camp.
  7. Are there restrictions on licenses of the to-be-released open source code?
    • An MIT license is recommended but can be changed with your mentor’s consent.
  8. Are there restrictions on the authorship of the to-be-released technical paper?
    • The paper should include you and your mentor but can be extended with your mentor’s consent.
  9. How does the mentoring work?
    • Most mentors are machine learning and/or deep learning experts, that have a full-time job. This means they will not be on-site most of the time, but are very enthusiatic to help you out! So please actively ask questions and start discussions with them.
  10. Can I develop on Keras or other ML framework/library?
    • You can use Keras including other framework/library that can run on top of TensorFlow.


Below are links to final presentations, from our fascinating participants:

Participant Title
Eyal Gruss Autotransformer
Masahiro Kaneko Document-level re-ranking for NMT
Hong Min Emotion-aware conversational interface [slides]
Ali Zaidi Multilingual transfer learning with transformer models
Brett Beaulieu-Jones Predicting healthcare outcomes with recurrent neural networks and concept embeddings
Anoop Toffy, Chaeyoung Lee Improving speech recognition accuracy using synthesised speech output from semi-supervised GANs
Leo Kim, Olga Slizovskaia Visually-informed music source separation [demo 1], [demo 2]
Feryal Behbahani Automated curriculum learning for reinforcement learning
Jeewoo Kim Competitive self-play on multi-snakes environment [slides]
Tegg Sung, Aleksandra Malysheva Multi-agents control using graph neural networks
Valliappa Chockalingam & Rishab Gargeya Evaluating generalization with distributional RL agents
Amlesh Sivanantham Imagination-augmented agents for deep reinforcement learning on solving Rubik’s cubes
Hyunsun Choi Uncertainty estimation in latent variable inference
Abien Fred Agarap On equilibrium of discriminative neural networks
Seo Yeon Stella Yang Deep learning aided GPS/INS smartphone pedestrian dead reckoning [slides]
Jaewook Kang Don’t be a turtle
Taekmin Kim Real-time video segmentation on mobile devices
Sein Jang Location aware recommendation on mobile devices [slides]
Oh-hyeon Choung Image semantic segmentation by integrating multi-scale feature maps
Carlos Santana Vega Long term video generation
Mudassar Liaq Unsupervised video summarization using reinforcement learning [slides]