The invention provides a video frame rate upconversion method and a video frame rate upconversion system based on a cyclic convolution neural network Dividing the initial video into groups of image blocks comprising two consecutive frames of images; taking two consecutive frames of images in the image block as inputs of a target cyclic convolution neural network, and synthesizing intermediate frame images corresponding to the two consecutive frames of images; wherein, the target cyclic convolution neural network is trained by a preset training data set The intermediate frame image is inserted into the image block to obtain the target video after the video frame rate upconversion Thus, the mapping from the front and back two frames to the middle frame can be completed, the frame rate ofthe original video is improved, and the up-conversion of the video frame rate is better completed

Video frame rate upconversion method and system based on cyclic convolution neural network

The invention provides a wallpaper splicing seam aligning device and a wallpaper splicing seam aligning method by utilizing the same. The device comprises an automatic guide transport vehicle, a firstintelligent camera, a second intelligent camera, a laser pen, an included angle fine adjustment mechanism and a displacement fine adjustment mechanism, wherein the automatic guide transport vehicle is provided with a wallpaper paving device, the first intelligent camera, the second intelligent camera and the laser pen are installed on the same side surface of a shell of the wallpaper paving device, moreover, the first intelligent camera, the second intelligent camera and the laser pen are perpendicular aligned, a lens of the first intelligent camera and a lens of the second intelligent cameraare opposite to the wall surface, the first intelligent camera and the second intelligent camera send control signals to the included angle fine adjustment mechanism and the displacement fine adjustment mechanism, the included angle fine adjustment mechanism is arranged on the automatic guide transportation vehicle, and is used for adjusting an included angle between the bottom surface of the wallpaper paving device and the horizontal plane under driving of the control signals, and the displacement fine adjustment mechanism is arranged on the automatic guide transportation vehicle, and is used for finely adjusting displacement of the wallpaper paving device in the direction parallel to the wall surface under driving of the control signal sent by the first intelligent camera and the secondintelligent camera.

Wallpaper splicing seam aligning device and wallpaper splicing seam aligning method by utilizing same

The invention provides a detection system and method for product surface defects. The system comprises a grasping mechanical hand, a detection mechanical hand, a control module and a defect detectiondevice, wherein the defect detection device comprises a first camera and a second camera; the first camera is used for shooting an image of the upper surface of a detected product; the second camera is used for shooting the image of the lower surface of the detected product; the first camera sends the collected image of the upper surface of the product to the control module; the control module detects whether the upper surface of the product has a defect or not according to the image of the upper surface of the product, and outputs a first detection result; the second camera sends the collected image of the lower surface of the product to the control module; and the control module detects whether the lower surface of the product has a defect or not according to the image of the lower surface of the product, and outputs a second detection result. Therefore, the automatic detection on the product surface defects can be realized.

Detection system and method for product surface defects

The invention discloses a steel surface defect recognition method and a device based on depth learning technology. The method comprises the following steps of: collecting corresponding steel surface defect pictures according to the type of steel production line; according to the labeling process of the labeling platform, the defects are labeled to generate the training sample set and the verification sample set. According to the types and characteristics of steel surface defects, a neural network model is designed. Training the neural network model according to the training sample set and thetraining flow of the neural network; evaluating each index of the neural network model according to the verification sample set and the evaluation method of the neural network model, and optimizing the neural network model; deployment and application of the neural network model are carried out in a distributed manner according to the device performance and the size of the neural network model. Byusing the depth learning technology and optimizing the neural network model, the equipment cost can be reduced, and the steel surface defects can be identified with real-time and high accuracy.

A method and a device for recognizing surface defects of steel based on depth learning technology

The invention discloses a universal object detection system and method for improving active learning based on a self-monitoring process. The system comprises: a sample obtaining unit, which obtains asmall number of labeled samples and a mass of unlabeled samples; a model establishment and initialization unit, which establishes the depth-learning object detection model and initializes the depth-learning object detection model with a small number of labeled samples; a self-monitoring learning unit, which introduces a self-supervised learning course to guide the self-supervised learning processto mine a large number of high confidence samples without labeling to automatically carry out pseudo-labeling, and carries out image cross-validation on the pseudo-labeled samples, carries out image mapping re-prediction, and selects pseudo-labeling samples with high prediction consistency by cross-voting; an active learning unit, which introduces the active learning course to guide the active learning process to mine the low confidence samples from the mass unlabeled samples for manual labeling; a model training unit, which is configured to train the selected pseudo-labeled samples with highprediction consistency and the manually labeled samples to improve the model performance.

Universal object detection system and method for improving active learning based on self-monitoring process

The invention discloses a convolutional neural network and active learning-based ceramic tile surface defect recognition method. The method comprises the following steps of: (1) obtaining and preprocessing an image; (2) establishing a training set; (3) establishing and training a convolutional neural network; (4) carrying out active learning; (5) carrying out model iteration; and (6) carrying outonline detection. Compared with the prior art, the method has the advantages that (1) features of ceramic tile surface defects are automatically extracted by using the convolutional neural network, sothat good priori knowledges are not required in the aspect of defect feature extraction, and multiple defect types in one to-be-detected image can be recognized; and (2) active learning is imported in convolutional neural network training, so that labeling cost of samples is effectively decreased and the model convergence is accelerated.

Yao Minghai

Papers

Convolutional neural network and active learning-based ceramic tile surface defect recognition method