Human Activity Recognition using Resnet-34 Model

Activity recognition has been an emerging field of research since the past few decades. Humans have ability to recognize activities from a number observations in their surroundings. These are used several areas like video surveillance, health sectors, gesture detection, energy conservation, fall detection systems and many more. Sensor based approaches accelerometer, gyroscope, etc., discussed with its advantages disadvantages. There different ways using sensors smartly controlled environment. A step-by-step procedure is followed this paper build human activity recognizer. general architecture Resnet model explained first along description workflow. Convolutional neural network which capable classifying trained kinetic dataset includes more than 400 classes activities. The videos last around tenth second. Resnet-34 for image classification convolutional networks it provides shortcut connections resolves problem vanishing gradient. tested successfully giving satisfactory result by recognizing over actions. Finally, some open problems presented should be addressed future research.