Deep convolutional neural networks have performed remarkably well on many Computer Vision tasks. However, these networks are heavily reliant on big data to avoid overfitting. Overfitting refers to the phenomenon when a network learns a function with very high variance such as to perfectly model the training data. Unfortunately, many application domains do not have access to big data, such as medical image analysis. This survey focuses on Data Augmentation, a data-space solution to the problem of limited data. Data Augmentation encompasses a suite of techniques that enhance the size and quality of training datasets such that better Deep Learning models can be built using them. The image augmentation algorithms discussed in this survey include geometric transformations, color space augmentations, kernel filters, mixing images, random erasing, feature space augmentation, adversarial training, generative adversarial networks, neural style transfer, and meta-learning. The application of augmentation methods based on GANs are heavily covered in this survey. In addition to augmentation techniques, this paper will briefly discuss other characteristics of Data Augmentation such as test-time augmentation, resolution impact, final dataset size, and curriculum learning. This survey will present existing methods for Data Augmentation, promising developments, and meta-level decisions for implementing Data Augmentation. Readers will understand how Data Augmentation can improve the performance of their models and expand limited datasets to take advantage of the capabilities of big data.

/pdf/a-survey-on-image-data-augmentation-for-deep-learning-4ru1bj2hbb.pdf

A survey on Image Data Augmentation for Deep Learning

Improved approaches for users to with graphical user interfaces of computing devices are disclosed. A rotational user action supplied by a user via a user input device can provide accelerated scrolling. The accelerated nature of the scrolling enables users to scroll or traverse a lengthy data set (e.g., list of items) faster and with greater ease. The amount of acceleration provided can be performed in successive stages, and/or performed based on the speed of the rotational user action. In one embodiment, the rotational user action is transformed into linear action with respect to a graphical user interface. The resulting acceleration effect causes the linear action to be enhanced such that a lengthy data set is able to be rapidly traversed.

Method and apparatus for accelerated scrolling

In accordance with some embodiments, a computer-implemented method for use in conjunction with a device with a touch screen display is disclosed. In the method, a movement of an object on or near the touch screen display is detected. In response to detecting the movement, an electronic document displayed on the touch screen display is translated in a first direction. If an edge of the electronic document is reached while translating the electronic document in the first direction while the object is still detected on or near the touch screen display, an area beyond the edge of the document is displayed. After the object is no longer detected on or near the touch screen display, the document is translated in a second direction until the area beyond the edge of the document is no longer displayed.

List scrolling and document translation, scaling, and rotation on a touch-screen display

The present disclosure discloses an object position detector. The object position detector comprises a touch sensor formed as a closed loop and having a physical constraint formed on an upper surface of the touch sensor and coextensive with the closed loop. The touch sensor is configured to sense motion of an object proximate to the closed loop. The object position detector also comprises a processor coupled to the touch sensor and is programmed to generate an action in response to the motion on the touch sensor.

Closed-loop sensor on a solid-state object position detector

An input device is disclosed. The input device includes a movable touch pad capable of detecting an object in close proximity thereto so as to generate a first control signal. The input device also includes a movement indicator capable of detecting the movements of the movable touch pad so as to generate one or more distinct second control signals. The control signals being used to perform actions in an electronic device operatively coupled to the input device.

Movable touch pad with added functionality

An electronic apparatus 20 is of a portable type and has a thin body integral with a display screen holdable by both hands HR, HL of a user for viewing purposes. Disposed on a front of the body 21 are a flat display panel 22 and various operation buttons 23–27 , and first and second touch sensors 28 and 29 . The first touch sensor 28 is disposed on the apparatus at a position where when the body 21 is held by both the hands of the user, a touch surface of the first sensor 28 is easily operated vertically by a finger of the user's right-hand HR whereas the second touch sensor 29 is disposed at a position where when the body 21 is held by both the hands of the user, its touch surface is easily operated horizontally by a finger of the user's left-hand HL.

Portable electronic apparatus and a display control method

This short paper reports the method and the evaluation results of Casio and Shinshu University joint team for the ISBI Challenge 2017 - Skin Lesion Analysis Towards Melanoma Detection - Part 3: Lesion Classification hosted by ISIC. Our online validation score was 0.958 with melanoma classifier AUC 0.924 and seborrheic keratosis classifier AUC 0.993.

Image Classification of Melanoma, Nevus and Seborrheic Keratosis by Deep Neural Network Ensemble.

A digital camera includes an image capturing unit, an image composition unit, and a display control unit. The image capturing unit captures frames at predetermined time intervals. The image composition unit sequentially combines at least a part of image data from image data of a plurality of frames sequentially captured by the image capturing unit at predetermined time intervals. The display control unit performs control to sequentially display image data combined by the image composition unit while the image data of the frames are captured by the image capturing unit at predetermined time intervals.

Image capturing apparatus capable of capturing panoramic image

An imaging apparatus includes: an imaging section configured to acquire a plurality of images that are consecutively taken in time series by an imaging device; a reference region setting section configured to employ one of the images acquired by the imaging section as a reference image and to set a region of the reference image as reference region, the region having high-frequency components; and an image processing section configured to combine the images including the reference image by positioning the images so as to overlap subject portions in each images, the subject portions being defined in the reference region set by the reference region setting section.

Imaging apparatus provided with panning mode for taking panned image

An image selection device including: a reception unit that receives a plurality of captured images obtained by rapid shooting one or more people as subjects; a face detection unit that detects human faces included in the captured images received by the reception unit; an eye detection unit that detects eyes in the human faces detected by the face detection unit; a blink detection unit that detects blink degrees of each of the eyes detected by the eye detection unit; an estimation unit that estimates statuses of the human faces based on the blink degrees detected by the blink detection unit; and a selecting unit that selects at least one of the captured images to be recorded on a recording medium from among the captured images received by the reception unit, based on the statuses of the human faces estimated by the estimation unit.

Kazuhisa Matsunaga

Papers

Portable electronic apparatus and a display control method

Image Classification of Melanoma, Nevus and Seborrheic Keratosis by Deep Neural Network Ensemble.

Image capturing apparatus capable of capturing panoramic image

Imaging apparatus provided with panning mode for taking panned image

Image selection device and method for selecting image