Abstract: An arithmetic operation method for a convolutional layer in a neural network includes: generating a coefficient matrix by converting a kernel used in the convolutional layer such that the coefficient matrix is associated with an input vector obtained by expanding, into one column, a feature map input to the convolutional layer; searching for non-zero elements included in the coefficient matrix; assigning multiplications of the non-zero elements included in the coefficient matrix and corresponding elements of the input vector to a plurality of calculators with each of the multiplications being handled as a unit of process, so as to level out the numbers of units of process among the calculators, each of the calculators being capable of performing a process in parallel with one another; and sequentially performing, by the calculators, the assigned multiplications, and sequentially adding, by the calculators, results of the multiplications to corresponding elements of an output vector.

Abstract: An operation circuit contains a parallel operation circuit including operators storage circuits. Each storage circuit has an input storage circuit that stores elements of an input vector in an order based on an index of each element; and a coefficient storage circuit including a ring buffer that stores elements of a row or column vector of a coefficient matrix in an order based on an index of each element. Each operator sequentially multiplies the elements of the coefficient matrix in the storage circuit by a corresponding one of the elements of the input vector, and adds a result of multiplication to a corresponding one of elements of the output vector based on the index thereof.

Abstract: An arithmetic operation method for a convolutional layer in a neural network includes: generating a coefficient matrix by converting a kernel used in the convolutional layer such that the coefficient matrix is associated with an input vector obtained by expanding, into one column, a feature map input to the convolutional layer; searching for non-zero elements included in the coefficient matrix; assigning multiplications of the non-zero elements included in the coefficient matrix and corresponding elements of the input vector to a plurality of calculators with each of the multiplications being handled as a unit of process, so as to level out the numbers of units of process among the calculators, each of the calculators being capable of performing a process in parallel with one another; and sequentially performing, by the calculators, the assigned multiplications, and sequentially adding, by the calculators, results of the multiplications to corresponding elements of an output vector.

Abstract: An operation circuit contains a parallel operation circuit including operators storage circuits. Each storage circuit has an input storage circuit that stores elements of an input vector in an order based on an index of each element; and a coefficient storage circuit including a ring buffer that stores elements of a row or column vector of a coefficient matrix in an order based on an index of each element. Each operator sequentially multiplies the elements of the coefficient matrix in the storage circuit by a corresponding one of the elements of the input vector, and adds a result of multiplication to a corresponding one of elements of the output vector based on the index thereof.

Abstract: A beam exposure device includes a light-emitting unit for emitting light beams from a plurality of light-emitting positions, a scan unit, an optical condensing system for condensing a spot of the light beams onto a surface to be exposed, and a micro-deflection unit for micro-deflecting the plurality of light beams to expose the space between the beams in the plurality of light beams. The optical condensing system includes a first microlens array arranged between the light-emitting unit and the micro-deflection unit and provided with a plurality of microlenses corresponding to the light-emitting positions; and a second microlens array arranged between the micro-deflection unit and the surface to be exposed and provided with a plurality of microlenses each microlens corresponding to the light-emitting unit.

Abstract: A beam exposure device includes a light-emitting unit for emitting light beams from a plurality of light-emitting positions, a scan unit, an optical condensing system for condensing a spot of the light beams onto a surface to be exposed, and a micro-deflection unit for micro-deflecting the plurality of light beams to expose the space between the beams in the plurality of light beams. The optical condensing system includes a first microlens array arranged between the light-emitting unit and the micro-deflection unit and provided with a plurality of microlenses corresponding to the light-emitting positions; and a second microlens array arranged between the micro-deflection unit and the surface to be exposed and provided with a plurality of microlenses each microlens corresponding to the light-emitting unit.

Abstract: This alignment device is furnished with an alignment light source for emitting alignment light, and is housed with a camera for example. The alignment light source emits alignment light, doing so, for example, coaxially with respect to the optical axis of light detected by the camera. The alignment light illuminates a substrate and mask, and reflected light is detected by the camera. A microlens array for exposure use is present between a mask alignment mark and a substrate alignment mark, whereby an erect unmagnified image reflected from the substrate alignment mark is formed on the mask. A control device then uses the mask alignment mark and the substrate alignment mark detected by the camera to perform alignment of the substrate and the mask. Alignment of the substrate and the mask can be performed with high accuracy thereby.

Abstract: A photo-alignment exposure method divides each unit image area of a liquid crystal display into a plurality of divided areas and photo-aligns an alignment material film of each of the divided areas in mutually different directions. The method includes a first exposure process that radiates light at an inclined photo-irradiation angle onto an exposed surface of entire the unit image area; and a second exposure process that radiates light onto one area of the divided areas at an inclined photo-irradiation angle different from the photo-irradiation angle in the first exposure process. In the second exposure process, the light is radiated through a mask pattern corresponding to one area of the divided areas, and transmitted light of the mask pattern is condensed by condensing element and radiated onto the area.

Abstract: A photo-alignment exposure device that includes a first mask and a first exposure device that independently proximity-exposes a first divided area, a second mask and a second exposure device that independently proximity-exposes a second divided area adjacent to the first divided area, and a third mask and a third exposure device that exposes an area on a side of the first divided area near a boundary between the first divided area and the second divided area. The third exposure device is provided with a photo-irradiation angle same as that of the first exposure device or the second exposure device with respect to an exposed surface. A condensing element that condenses the mask transmitted light on the area on a side of the first divided area near the boundary is provided between the mask opening of the third mask and the exposed surface.

Abstract: Exposure apparatus includes photomasks on which a mask pattern having the same shape as that of an exposure pattern exposed onto a surface of a TFT substrate held on a stage is formed, lens assemblies in which unit lens groups in each of which a plurality of convex lenses are arranged in a normal direction to the photomasks so that same-size erect images of mask patterns formed on the photomasks can be formed on the surface of the TFT substrate are arranged in a plane parallel with the photomasks and the surface of the TFT substrate held on the stage, and moving device that moves the lens assemblies in a plane parallel with the masks and the surface of TFT substrate held on the stage.

Abstract: A photo-alignment exposure method divides each unit image area of a liquid crystal display into a plurality of divided areas and photo-aligns an alignment material film of each of the divided areas in mutually different directions. The method includes a first exposure process that radiates light at an inclined photo-irradiation angle onto an exposed surface of entire the unit image area; and a second exposure process that radiates light onto one area of the divided areas at an inclined photo-irradiation angle different from the photo-irradiation angle in the first exposure process. In the second exposure process, the light is radiated through a mask pattern corresponding to one area of the divided areas, and transmitted light of the mask pattern is condensed by condensing element and radiated onto the area.

Abstract: A photo-alignment exposure device that includes a first mask and a first exposure device that independently proximity-exposes a first divided area, a second mask and a second exposure device that independently proximity-exposes a second divided area adjacent to the first divided area, and a third mask and a third exposure device that exposes an area on a side of the first divided area near a boundary between the first divided area and the second divided area. The third exposure device is provided with a photo-irradiation angle same as that of the first exposure device or the second exposure device with respect to an exposed surface. A condensing element that condenses the mask transmitted light on the area on a side of the first divided area near the boundary is provided between the mask opening of the third mask and the exposed surface.

Abstract: On a film where an exposure material coating has been formed in a exposure pattern formation region on a film base material, a colored firing material, colored light-curable material, or colored ink is applied to at least one of two widthwise side edges to form a side part application coating, which is irradiated with laser light by an alignment mark formation unit to form an alignment mark. The alignment mark is then used to detect film meandering and adjust the positions of masks. This makes it easy to form the alignment mark and detect the alignment mark thus formed and makes it possible to accurately correct for meandering of a film and stably expose the film in the process of continuous exposure of a film where an exposure material coating has been formed in a exposure pattern formation region on a film base material.

Abstract: This alignment device is furnished with an alignment light source for emitting alignment light, and is housed with a camera for example. The alignment light source emits alignment light, doing so, for example, coaxially with respect to the optical axis of light detected by the camera. The alignment light illuminates a substrate and mask, and reflected light is detected by the camera. A microlens array for exposure use is present between a mask alignment mark and a substrate alignment mark, whereby an erect unmagnified image reflected from the substrate alignment mark is formed on the mask. A control device then uses the mask alignment mark and the substrate alignment mark detected by the camera to perform alignment of the substrate and the mask. Alignment of the substrate and the mask can be performed with high accuracy thereby.

Abstract: On a film where an exposure material coating has been formed in a exposure pattern formation region on a film base material, a colored firing material, colored light-curable material, or colored ink is applied to at least one of two widthwise side edges to form a side part application coating, which is irradiated with laser light by an alignment mark formation unit to form an alignment mark. The alignment mark is then used to detect film meandering and adjust the positions of masks. This makes it easy to form the alignment mark and detect the alignment mark thus formed and makes it possible to accurately correct for meandering of a film and stably expose the film in the process of continuous exposure of a film where an exposure material coating has been formed in a exposure pattern formation region on a film base material.

Abstract: One aspect of the invention provides a liquid crystal display device producing method for irradiating a liquid crystal display substrate, in which plural pixels are formed in a matrix state and liquid crystal is sealed between a TFT substrate and a counter electrode substrate, with light having a predetermined wavelength to orient liquid crystal molecules toward a predetermined direction in a state in which an electric field is applied to each pixel of the liquid crystal display substrate. The method includes the steps of: dipping the liquid crystal display substrate and a lamp in a transparent liquid having resistivity of a predetermined value or more and sufficiently high transmittance to the light in a state in which the liquid crystal display substrate and the lamp face each other; and lighting the lamp to irradiate the liquid crystal display substrate with the light having a predetermined light quantity in a state in which the electric field is applied to each pixel.

Abstract: The present invention is a photomask 3 for exposing a substrate coated with a positive photosensitive material. At least a first mask pattern group 16 and a second mask pattern group 17 are formed on a transparent substrate at a predetermined arrangement pitch. The first mask pattern group 16 has first light shielding patterns 20 arranged at an interval corresponding to two types of convex pattern forming portions of different heights on the substrate, in which the first light shielding patterns 20 each have a substantially same area as a cross sectional area of a convex pattern. The second mask pattern group 17 has a second light shielding pattern 22 and an opening pattern 23, in which the second light shielding pattern 22 has a predetermined area and corresponds to a higher convex pattern forming portion among the two types of convex pattern forming portions, and the opening pattern corresponds to a lower convex pattern forming portion.

Abstract: Exposure apparatus includes photomasks on which a mask pattern having the same shape as that of an exposure pattern exposed onto a surface of a TFT substrate held on a stage is formed, lens assemblies in which unit lens groups in each of which a plurality of convex lenses are arranged in a normal direction to the photomasks so that same-size erect images of mask patterns formed on the photomasks can be formed on the surface of the TFT substrate are arranged in a plane parallel with the photomasks and the surface of the TFT substrate held on the stage, and moving device that moves the lens assemblies in a plane parallel with the masks and the surface of TFT substrate held on the stage.

Abstract: The present invention is a photomask 3 for exposing a substrate coated with a positive photosensitive material. At least a first mask pattern group 16 and a second mask pattern group 17 are formed on a transparent substrate at a predetermined arrangement pitch. The first mask pattern group 16 has first light shielding patterns 20 arranged at an interval corresponding to two types of convex pattern forming portions of different heights on the substrate, in which the first light shielding patterns 20 each have a substantially same area as a cross sectional area of a convex pattern. The second mask pattern group 17 has a second light shielding pattern 22 and an opening pattern 23, in which the second light shielding pattern 22 has a predetermined area and corresponds to a higher convex pattern forming portion among the two types of convex pattern forming portions, and the opening pattern corresponds to a lower convex pattern forming portion.

Abstract: One aspect of the invention provides a liquid crystal display device producing method for irradiating a liquid crystal display substrate, in which plural pixels are formed in a matrix state and liquid crystal is sealed between a TFT substrate and a counter electrode substrate, with light having a predetermined wavelength to orient liquid crystal molecules toward a predetermined direction in a state in which an electric field is applied to each pixel of the liquid crystal display substrate. The method includes the steps of: dipping the liquid crystal display substrate and a lamp in a transparent liquid having resistivity of a predetermined value or more and sufficiently high transmittance to the light in a state in which the liquid crystal display substrate and the lamp face each other; and lighting the lamp to irradiate the liquid crystal display substrate with the light having a predetermined light quantity in a state in which the electric field is applied to each pixel.