azzaelnaggar / action_classification

Video Action Classification

about project:

video classification using the UCF101 and kinetics-700 action recognition dataset.

dataset_links

https://www.deepmind.com/open-source/kinetics

https://www.tensorflow.org/datasets/catalog/ucf101

Selected classes from ucf101 are

['ApplyEyeMakeup','ApplyLipstick','Archery','BabyCrawling','BalanceBeam','BandMarching', 'BaseballPitch','BasketballDunk','Basketball','BenchPress','Biking', 'Billiards','BlowDryHair','BlowingCandles','BodyWeightSquats','Bowling', 'BoxingPunchingBag','BoxingSpeedBag','BreastStroke','BrushingTeeth']

• We used the 2 datasets to make a prediction but used ucf101 for production.

• We make many traials to get the final results.

• Deploy tensorflow model using flask.

OUTPUT