id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
work_dj5mnfkkk5dh3oloxxwch56axi	Chanuki Illushka Seresinhe	Using deep learning to quantify the beauty of outdoor places	2017	15	.pdf	application/pdf	8874	1919	64	use the Places CNN [32,33] to extract hundreds of features from over 200 000 outdoor images from We explore data extracted from images from Scenic-Or-Not, an online game that crowdsources ratings we use the Scenic-Or-Not dataset to understand what characteristics beautiful images of our environment and 'Open Area' are both negatively associated with scenicness in our model containing all images (b) Unscenic images appear to be mainly composed of man-made features, e.g. industrial areas, road networks, However, the most scenic images in urban built-up areas can also include man-made features such trained on urban built-up areas (depicted here) and the model trained on all of our Scenic-Or-Not images (depicted in figure 2), large we check whether we can use a CNN to predict the scenic ratings of images with a high degree of accuracy. Scenic CNN in general predicts low ratings for images containing primarily man-made features, images Our Scenic CNN predicts high ratings for images containing	./cache/work_dj5mnfkkk5dh3oloxxwch56axi.pdf	./txt/work_dj5mnfkkk5dh3oloxxwch56axi.txt