id	author	title	date	pages	extension	mime	words	sentence	flesch	summary	cache	txt
fcis-12004	Yu, Zefeng	Research on Video Detection Method of Mudslide based on Inflated 3D Convolutional Neural Network	2023	4	.pdf	application/pdf	3522	161	49	Also, because the video captured by the camera has a certain degree of noise as well as light and shadow variations that can cause a certain amount of error in the results of the data, this paper uses the I3D training model to extract video features and the extracted features are used to calculate the anomaly score using the Sigmoid activation function, and based on the anomaly score, it is predicted whether or not an anomalous event occurs. 3.3. Next, the input is a video clip consisting of a series of consecutive video frames that are closely connected in time and represent video content over a period of time.	cache/fcis-12004.pdf	txt/fcis-12004.txt