以視覺資訊預估天氣屬性;Visual Weather Property Prediction

[[abstract]]在本論文中，我們嘗試單純使用圖片的資訊，利用卷積遞迴神經網路來預估天氣屬性。我們研究了兩種情況下的天氣屬性預估:1)預估單張室外的照片溫度2)預測一序列照片中，最後一張照片的天氣屬性。第一種情況下，使用以大規模影像數據庫訓練的卷積神經網路提取視覺特徵，我們證明可獲得相當好的性能，並分析訓練數據量如何影響性能。在第二種情況下，我們考慮視覺外觀隨著時間的變化，建構出一個遞迴神經網路來預測一序列圖片中最後一張圖片的天氣屬性。與目前最先進的模型相比，我們獲得了更好的預測準確度。此外，我們研究從場景不同區域提取視覺資訊以及在不同的白天時間拍攝的照片對性能的影響。我們的方法進一步的強化僅使用圖片中視覺資訊來進行有效天氣屬性預測的這個想法。 In thesis paper, we attempt to employ convolutional recurrent neural networks for weather weather properties estimation using only image data. We study ambient weather properties estimation based on deep neural networks in two scenarios: a) estimating temperature of a single outdoor image, and b) predicting weather properties of the last image in an image sequence. In the first scenario, visual features are extracted by a convolutional neural network trained on a large-scale image dataset. We demonstrate that promising performance can be obtained, and analyze how volume of training data influences performance. In the second scenario, we consider the temporal evolution of visual appearance, and construct a recurrent neural network to predict weather properties of the last image in a given image sequence. We obtain better prediction accuracy compared to the state-of-the-art models. Further, we investigate how performance varies when information is extracted from different scene regions, and when images are captured in different daytime hours. Our approach further reinforces the idea of using only visual information for cost efficient weather prediction in the future