This was my second time participating in the Kaggle competition, TensorFlow - Help Protect the Great Barrier Reef, and the deadline was only 1 month to go. The topic domain of the competition was about object detection of computer vision.
As you can imagine, most people and teams were using the architecture of the YOLO series to train the models, and most of the discussions were about YOLO methods for optimizations.
Honestly, I think that this time the dataset was not distributed really well indeed. Many teams were ranked vary significantly between public LB and private LB.
Besides, in my opinion, I really thought that it is highly difficult to get a good result (or score/rank) if you do not have enough (awesome) hardware resources.
In short, I appreciate my teammates who have made a lot of contributions as well. I am so glad that we can win the silver medal from over 2000 teams on the public LB of this competition. 😃