Abstract: Disclosed herein is a system and method for improving the accuracy of an object detector when trained with a dataset having a significant number of missing annotations. The method uses a novel Background Recalibration Loss (BRL) which adjusts the gradient direction according to its own activation to reduce the adverse effect of error signals by replacing the negative branch of the focal loss with a mirror of the positive branch when the activation is below a confusion threshold.

Abstract: The present disclosure provides a back plate for use in a display apparatus, wherein the back plate includes: a bottom wall; a through hole in the bottom wall, wherein the through hole is provided with a hole wall protruding towards a side of the bottom wall, and the hole wall and the bottom wall are connected together and have a one-piece structure, wherein materials of the bottom wall and the hole wall each include an aluminum alloy material. The present disclosure further provides a manufacturing method of the back plate, a backlight module and a display apparatus.

Abstract: Backlight module and display device are disclosed. The backlight module includes a light source structure and a light guide plate, which includes a body structure and a wedge-shaped structure disposed on the body structure and located at a side adjacent to the light source structure. The body structure includes a light incident surface of body structure. The wedge-shaped structure includes a bottom surface, a wedge surface and a light incident surface of wedge-shaped structure. The light source structure includes a light source, light emitted by which enters the light guide plate through the light incident surface of body structure and the light incident surface of wedge-shaped structure. In a direction perpendicular to the bottom surface, a ratio of thickness of the body structure to thickness of light-exiting surface of the light source is less than or equal to 88%.

Abstract: A method of detecting target object includes: extracting, through a neural network, a feature of a reference frame and a feature of a frame under detection; inputting each of at least two feature groups from at least two network layers of the neural network into a detector so as to obtain a corresponding detection result group output from the detector; wherein each feature group includes features of the reference frame and of the frame under detection, each detection result group includes a classification result and a regression result with respect to each of a plurality of candidate boxes for a feature group; and acquiring a bounding box for the target object in the frame under detection according to the at least two detection result groups.

Abstract: A method of detecting target object includes: extracting, through a neural network, a feature of a reference frame and a feature of a frame under detection; inputting each of at least two feature groups from at least two network layers of the neural network into a detector so as to obtain a corresponding detection result group output from the detector; wherein each feature group includes features of the reference frame and of the frame under detection, each detection result group includes a classification result and a regression result with respect to each of a plurality of candidate boxes for a feature group; and acquiring a bounding box for the target object in the frame under detection according to the at least two detection result groups.