Narrow Results

As the surveillance devices proliferate, various machine learning approaches for video anomaly detection have been attempted. We propose a hybrid deep learning model composed of a video feature extractor trained by generative adversarial network with deficient anomaly data and an anomaly detector boosted by transferring the extractor. Experiments with UCSD pedestrian dataset show that it achieves 94.4% recall and 86.4% precision, which is the competitive performance in video anomaly detection.

The contradiction between the supply and demand of water resources is becoming increasingly prominent, whose main reason is the eutrophication of rivers and lakes. However, limited and inaccurate data makes it impossible to establish a precise model to successfully predict eutrophication levels. Moreover, it is incompetent to distinguish the degree of eutrophication status of lakes by manual calculation and processing. Focusing on these inconveniences, this study proposes 3D fractal net CNN to extract features in remote sensing images automatically, aiming at achieving scientific forecasting on eutrophication status of lakes. In order to certificate the effectiveness of the proposed method, we predict the state of the water body based on remote sensing images of natural lake. The images in natural lake were accessed by MODIS satellite, cloud-free chlorophyll inversion picture of 2009 was resized into $273\times 273$ patches, which were collected as training and testing samples. In the total of 162 pictures, our study makes three consecutive pictures as a set of data so as to attain 120 group of training and 40 testing data. Taking one set of data as input of the neural network and the next day’s eutrophication level as labels, CNNs act considerable efficiency. Through the experimental results of 2D CNN, 3D CNN and 3D fractal net CNN, 3D fractal net CNN has more outstanding performance than the other two, with the prediction accuracy of 67.5% better than 47.5% and 62.5%, respectively.

Developing systems for the automatic detection of events in video is a task which has gained attention in many areas including sports. More specifically, event detection for soccer videos has been studied widely in the literature. However, there are still a number of shortcomings in the state-of-the-art such as high latency, making it challenging to operate at the live edge. In this paper, we present an algorithm to detect events in soccer videos in real time, using 3D convolutional neural networks. We test our algorithm on three different datasets from SoccerNet, the Swedish Allsvenskan, and the Norwegian Eliteserien. Overall, the results show that we can detect events with high recall, low latency, and accurate time estimation. The trade-off is a slightly lower precision compared to the current state-of-the-art, which has higher latency and performs better when a less accurate time estimation can be accepted. In addition to the presented algorithm, we perform an extensive ablation study on how the different parts of the training pipeline affect the final results.