No abstract

Using convolutional neural networks for 360 • images can induce sub-optimal performance due to distortions entailed by a planar projection. The distortion gets deteriorated when a rotation is applied to the 360 • image. Thus, many researches based on convolutions attempt to reduce the distortions to learn accurate representation. In contrast, we leverage the transformer architecture to solve image classification problems for 360 • images. Using the proposed transformer for 360 • images has two advantages. First, our method does not require the erroneous planar projection process by sampling pixels from the sphere surface. Second, our sampling method based on regular polyhedrons makes low rotation equivariance errors, because specific rotations can be reduced to permutations of faces. In experiments, we validate our network on two aspects, as follows. First, we show that using a transformer with highly uniform sampling methods can help reduce the distortion. Second, we demonstrate that the transformer architecture can achieve rotation equivariance on specific rotations. We compare our method to other state-of-the-art algorithms using the SPH-MNIST, SPH-CIFAR, and SUN360 datasets and show that our method is competitive with other methods.

We present a novel method for the upright adjustment of 360 • images. Our network consists of two modules, which are a convolutional neural network (CNN) and a graph convolutional network (GCN). The input 360 • images is processed with the CNN for visual feature extraction, and the extracted feature map is converted into a graph that finds a spherical representation of the input. We also introduce a novel loss function to address the issue of discrete probability distributions defined on the surface of a sphere. Experimental results demonstrate that our method outperforms fully connectedbased methods.

Drones became popular video capturing tools. Drone videos in the wild are first captured and then edited by humans to contain aesthetically pleasing camera motions and scenes. Therefore, edited drone videos have extremely useful information for cinematography and for applications such as camera path planning to capture aesthetically pleasing shots. To design intelligent camera path planners, learning drone camera motions from these edited videos is essential. However, first, this requires to filter drone clips and extract their camera motions out of these edited videos that commonly contain both drone and non‐drone content. Moreover, existing video search engines return the whole edited video as a semantic search result and cannot return only drone clips inside an edited video. To address this problem, we proposed the first approach that can automatically retrieve drone clips from an unlabeled video collection using high‐level search queries, such as “drone clips captured outdoor in daytime from rural places”. The retrieved clips also contain camera motions, camera view, and 3D reconstruction of a scene that can help develop intelligent camera path planners. To train our approach, we needed numerous examples of edited drone videos. To this end, we introduced the first large‐scale dataset composed of edited drone videos. This dataset is also used for training and validating our drone video filtering algorithm. Both quantitative and qualitative evaluations have confirmed the validity of our method.