Abstract: A calculation scheme decision system includes a pre-calculation unit performing, in an execution environment in which calculation is performed, calculation for each of respective layers of the network structure using at least one of calculation schemes prepared in advance for the respective layers, a cost acquisition unit acquiring a calculation cost of at least one calculation scheme for each layer based on a result of the calculation by the pre-calculation unit, a decision unit selecting one calculation scheme for each layer based on the calculation cost from among at least one of the calculation schemes prepared in advance for the respective layers to associate the layer with the selected one calculation scheme, and a calculation unit performing the calculation for each of the respective layers of the network structure on input data in the execution environment using the calculation scheme associated with each layer.

Abstract: A training system comprises a training device, and a training data creation device for the training device. The training device trains a neural network by means of a backpropagation algorithm. The training data creation device acquires any one of a positive evaluation indicating that content of the input data coincides with the label, a negative evaluation indicating that content of the input data does not coincide with the label, and an ignorable evaluation indicating exclusion from a training target label, for each label regarding the input data to create training data.

Abstract: A training system comprises a training device, and a training data creation device. The training device trains a neural network. The training data creation device acquires any one of a positive evaluation indicating that content of the input data coincides with the label, a negative evaluation indicating that content of the input data does not coincide with the label, and an ignorable evaluation indicating exclusion from a training target label, for each label regarding the input data to create training data. In training the neural network for training, the training system adjusts the weight coefficient for the intermediate layer and makes the recognition score of the label with the ignorable evaluation not affect the adjustment of the weight coefficient for the intermediate layer.

Abstract: A calculation scheme decision system includes a pre-calculation unit performing, in an execution environment in which calculation is performed, calculation for each of respective layers of the network structure using at least one of calculation schemes prepared in advance for the respective layers, a cost acquisition unit acquiring a calculation cost of at least one calculation scheme for each layer based on a result of the calculation by the pre-calculation unit, a decision unit selecting one calculation scheme for each layer based on the calculation cost from among at least one of the calculation schemes prepared in advance for the respective layers to associate the layer with the selected one calculation scheme, and a calculation unit performing the calculation for each of the respective layers of the network structure on input data in the execution environment using the calculation scheme associated with each layer.

Abstract: An image processing device for stitching a plurality of input images together so as to generate a panoramic composite image is provided. An imaging control section sets a guide for photographing a photographing subject from different imaging directions as a guide information for obtaining a plurality of input images suitable for the image composition. A guide display setting section displays a photographing guide image based on the guide information. A user rotates a camera while checking the guide image and captures the plurality of input images from different imaging directions. An image compositing section stitches the plurality of input images together so as to generate a composite image.

Abstract: The present disclosure can obtain a high-quality composite image by determining a suitable boundary when input images are sequentially stitched together to generate the composite image. An image generating apparatus 1 generates a composite image by stitching sequentially input images together. A difference value calculating section 15 calculates difference values by using the pixel values of a reference image and the pixel values of a target image that partly overlap the reference image. The difference values represent the relative relationship between the reference image and the target image. A boundary determining section 16 determines a boundary for stitching the reference image and the target image together by using the difference values calculated by the difference value calculating section 15. Then, an image compositing section 18 generates the composite image by stitching the reference image and the target image together based on the boundary determined by the boundary determining section 16.

Abstract: (Problem) To facilitate setting of an image display style and selection of images in an image display apparatus with a touch screen. (Solution to Problem) Thumbnails included in an area designated by a pinch gesture are all specified. If it is not true that “the thumbnails included in the designated area are all displayed in the selected style” (S2004, No), the thumbnails included in the designated area are all displayed in the selected style (S2005), and the status is set to selected state (S2006). If it is true that “the thumbnails included in the designated area are all displayed in the selected style” (S2004, Yes), the thumbnails included in the designated area are all displayed in unselected style (S2007), and the status is set to unselected state (S2008). If the gesture is continuously changed to a pinch-out gesture or a pinch-in gesture (S2010, No), the selected area or the unselected area are expanded or reduced depending on the status (S2014, S2015).

Abstract: An image processing device includes an input unit, a selection unit, a matching unit, an estimation unit and a merging unit. The input unit sequentially inputs images. The selection unit selects a reference image from among the input images. The matching unit calculates a correspondence relationship between a feature point of the reference image and a feature point of a target image. The estimation unit estimates an equation that causes a coordinate system of the reference image to correspond to a coordinate system of the target image using positional information of a pair of the feature points whose correspondence relationship has been calculated by the matching unit by regarding a motion between the reference image and the target image as only a rotational motion of the imaging element. The merging unit merges the input image and the target image to generate the composite image based on the equation.

Abstract: A training system comprises a training device, and a training data creation device. The training device trains a neural network. The training data creation device acquires any one of a positive evaluation indicating that content of the input data coincides with the label, a negative evaluation indicating that content of the input data does not coincide with the label, and an ignorable evaluation indicating exclusion from a training target label, for each label regarding the input data to create training data. In training the neural network for training, the training system adjusts the weight coefficient for the intermediate layer and makes the recognition score of the label with the ignorable evaluation not affect the adjustment of the weight coefficient for the intermediate layer.

Abstract: The present disclosure can obtain a high-quality composite image by determining a suitable boundary when input images are sequentially stitched together to generate the composite image. An image generating apparatus 1 generates a composite image by stitching sequentially input images together. A difference value calculating section 15 calculates difference values by using the pixel values of a reference image and the pixel values of a target image that partly overlap the reference image. The difference values represent the relative relationship between the reference image and the target image. A boundary determining section 16 determines a boundary for stitching the reference image and the target image together by using the difference values calculated by the difference value calculating section 15. Then, an image compositing section 18 generates the composite image by stitching the reference image and the target image together based on the boundary determined by the boundary determining section 16.

Abstract: A method and apparatus for partially up/downscaling an image encoded on a macroblock basis utilizing a computer. The method and apparatus performs operations of: creating map data of the encoded image so as to be able to find a filestream position corresponding to a macroblock in the encoded image to be decoded; creating a shrunken image of the encoded image; receiving a first control request from a user to display a first part in the encoded image; displaying on a display device, a second part in the shrunken image corresponding to the first part in the encoded image; decoding macroblocks containing the first part in the encoded image, using the map data; and displaying on the display device, the first part in the encoded image using the decoded macroblocks so as to replace the second part in the shrunken image therewith.

Abstract: The present disclosure can obtain a high-quality composite image by determining a suitable boundary when input images are sequentially stitched together to generate the composite image. An image generating apparatus 1 generates a composite image by stitching sequentially input images together. A difference value calculating section 15 calculates difference values by using the pixel values of a reference image and the pixel values of a target image that partly overlap the reference image. The difference values represent the relative relationship between the reference image and the target image. A boundary determining section 16 determines a boundary for stitching the reference image and the target image together by using the difference values calculated by the difference value calculating section 15. Then, an image compositing section 18 generates the composite image by stitching the reference image and the target image together based on the boundary determined by the boundary determining section 16.

Abstract: An image processing device for stitching a plurality of input images together so as to generate a panoramic composite image is provided. An imaging control section sets a guide for photographing a photographing subject from different imaging directions as a guide information for obtaining a plurality of input images suitable for the image composition. A guide display setting section displays a photographing guide image based on the guide information. A user rotates a camera while checking the guide image and captures the plurality of input images from different imaging directions. An image compositing section stitches the plurality of input images together so as to generate a composite image.

Abstract: A method and apparatus for partially up/downscaling an image encoded on a macroblock basis. The method and apparatus performs operations of: storing the encoded image; creating map data from bitstream of the encoded image to decode at a least one macroblock of the encoded image, creating a shrunken image of a predetermined size based on resolution of a display device, storing the map data and the shrunken image so as to relate with the encoded image; outputting the shrunken image related with the encoded image to be displayed based on a control request received from an input device; determining at least one macroblock to be decoded based on a display area of the shrunken image; partially decoding the encoded image for the determined macroblock using the map data; and outputting to the display device, the image data of the display area of the partially decoded image.

Abstract: The present disclosure can obtain a high-quality composite image by determining a suitable boundary when input images are sequentially stitched together to generate the composite image. An image generating apparatus 1 generates a composite image by stitching sequentially input images together. A difference value calculating section 15 calculates difference values by using the pixel values of a reference image and the pixel values of a target image that partly overlap the reference image. The difference values represent the relative relationship between the reference image and the target image. A boundary determining section 16 determines a boundary for stitching the reference image and the target image together by using the difference values calculated by the difference value calculating section 15. Then, an image compositing section 18 generates the composite image by stitching the reference image and the target image together based on the boundary determined by the boundary determining section 16.

Abstract: A method and apparatus for partially up/downscaling an image encoded on a macroblock basis. The method and apparatus performs operations of: storing the encoded image; creating map data from bitstream of the encoded image to decode at a least one macroblock of the encoded image, creating a shrunken image of a predetermined size based on resolution of a display device, storing the map data and the shrunken image so as to relate with the encoded image; outputting the shrunken image related with the encoded image to be displayed based on a control request received from an input device; determining at least one macroblock to be decoded based on a display area of the shrunken image; partially decoding the encoded image for the determined macroblock using the map data; and outputting to the display device, the image data of the display area of the partially decoded image.

Abstract: A method and apparatus for partially up/downscaling an image encoded on a macroblock basis utilizing a computer. The method and apparatus performs operations of: creating map data of the encoded image so as to be able to find a filestream position corresponding to a macroblock in the encoded image to be decoded; creating a shrunken image of the encoded image; receiving a first control request from a user to display a first part in the encoded image; displaying on a display device, a second part in the shrunken image corresponding to the first part in the encoded image; decoding macroblocks containing the first part in the encoded image, using the map data; and displaying on the display device, the first part in the encoded image using the decoded macroblocks so as to replace the second part in the shrunken image therewith.

Abstract: A method and apparatus for partially up/downscaling an image encoded on a macroblock basis. The method and apparatus performs operations of: storing the encoded image; creating map data from bitstream of the encoded image to decode at least one macroblock of the encoded image, creating a shrunken image of a predetermined size based on resolution of a display device, storing the map data and the shrunken image so as to relate with the encoded image; outputting the shrunken image related with the encoded image to be displayed based on a control request received from an input device; determining at least one macroblock to be decoded based on a display area of the shrunken image; partially decoding the encoded image for the determined macroblock using the map data; and outputting to the display device, the image data of the display area of the partially decoded image.

Abstract: Thumbnails included in an area designated by a pinch gesture are all specified. If it is not true that “the thumbnails included in the designated area are all displayed in the selected style” (S2004, No), the thumbnails included in the designated area are all displayed in the selected style (S2005), and the status is set to selected state (S2006). If it is true that “the thumbnails included in the designated area are all displayed in the selected style” (S2004, Yes), the thumbnails included in the designated area are all displayed in unselected style (S2007), and the status is set to unselected state (S2008). If the gesture is continuously changed to a pinch-out gesture or a pinch-in gesture (S2010, No), the selected area or the unselected area are expanded or reduced depending on the status (S2014, S2015).

Abstract: An image processing device includes an input unit, a selection unit, a matching unit, an estimation unit and a merging unit. The input unit sequentially inputs images. The selection unit selects a reference image from among the input images. The matching unit calculates a correspondence relationship between a feature point of the reference image and a feature point of a target image. The estimation unit estimates an equation that causes a coordinate system of the reference image to correspond to a coordinate system of the target image using positional information of a pair of the feature points whose correspondence relationship has been calculated by the matching unit by regarding a motion between the reference image and the target image as only a rotational motion of the imaging element. The merging unit merges the input image and the target image to generate the composite image based on the equation.