When solving the camera scene detection problem, one of the most critical challenges is to get high-quality diverse data for training the model. Since no public datasets existed for this task, a new Camera Scene Detection Dataset (CamSDD) containing more than 11K images and consisting of 30 different categories was collected first. The photos were crawled from Flickr and inspected manually to remove monochrome and heavily edited pictures, images with distorted colors and watermarks, photos that are impossible for smartphone cameras (e.g., professional underwater or night shots), etc. The dataset was designed to contain diverse images, therefore each scene category contains photos taken in different places, from different viewpoints and angles: e.g., the 'cat' category does not only contain cat faces but also normal full-body pictures shot from different positions. This diversity is essential for training a model that is generalizable to different environments and shooting conditions. Each image from the CamSDD dataset belongs to only one scene category. The dataset was designed to be balanced, thus each category contains on average around 350 photos. After the images were collected, they were resized to 576 x 384 px resolution as using larger photos will not bring any information that is vital for the considered classification problem.

Download  Camera  Scene  Detection  Dataset  (CamSDD)

While the proposed models demonstrate high accuracy on the CamSDD dataset, their real performance on live camera data is the most important for this task. For this, we developed an Android application that is using the obtained TensorFlow Lite models to perform real-time classification of the image frames coming from camera stream. We checked the predictions of the models on hundreds of different scenery, some sample results obtained with the Samsung Galaxy J5 are provided below.