# 7th place brief solution summary

Congratulation to all of the winners!
Most of the top solutions are based on 2.5D/3D classification of events from full image. For some reason I didn't believe this kind-of-simple pipeline would benefit and score that high, therefore I decided to stick to multiple stage approach.
To be short:
Detect ball (and throwiner)
Track ball
Classify area around the ball
No ensemble (solo detector + solo classifier). I used all the frames from videos in original resolution (FullHD)
To be more specific:
I used images from clips and manually annotated them (around 6k images). Firstly it was only ball detection, but I was successful only with classifying 'play' events, while throw-ins and challenges struggled. Then I added 2nd class of throwiner to my detector. It helped me to boost throwins classification further. I also used synthetic augmentation for throwiner class (removed actual throwiner, applied image inpainting to fill empty background, and added throwiner from another image instead).
I used Kalman-IoU tracker to keep the ball position visible even if it disappears. Unfortunately I couldn't tune params too well, and the ball position was unclear during aerial trajectories with high speed and acceleration of the ball.
I used 3D classifier with 4 classes (including background). I stacked center frame +- 12 frames, and classified overall 25 frames.
What helped me to improve the score:
Changing from 2.5D to 3D. Many tested 2.5D nets were worse than simple x3d-m.
Introducing throwiner class for detection. I had 0.45-0.5 accuracy for throwin events when I used only 3d classification. Adding throwiner class for detector gave me huge boost for throwin events, resulting in 0.8+ accuracy for throwin events. I merged 3d classification results and throwin class from detector.
Synthetic data generation for detector resulted in better stability for throwiner and ball near the line detection.
I was also struggling with time limits, since I was always close to 9 hours. Using FileVideoStream from imutils helped me to save some time, since it worked faster than cv2 video reader.
Unfortunately I couldn't improve challenges events. The score was always about 0.5-0.6 for it, while play and throwins scored 0.8-0.85 always. The main reason I believe is that I was losing ball due to its fast displacement during aerial and other challenges. My tracker couldn't solve that issue.