Hon Hai Precision Industry Co., Ltd.

Jung-Hao Yang

Chin-Pin Kuo

Chih-Te Lu

Tung-Tso Tsai

Tzu-Chen Lin

Chih-Yuan Chuang

Wan-Jhen Lee

Wei-chun Wang

Tsung-Yuan Tu

Guo-Chin Sun

Chung-yu Wu

Chien-Tsung Lee

A method acquires a first image and a second image of a target object being inputted into the depth estimation network for outputting a depth image. A pixel posture conversion relationship between the first image and the second image is obtained. The pixel posture conversion relationship includes a position relationship between each first pixel in the first image and a second pixel in the second image, which correspond to a same part of the target object. A restored image is generated based on the depth image, the pixel posture conversion relationship, and pre-obtained camera parameters. A loss of the depth estimation network is determined based on a difference between the first image, the depth image, the restored image, and the second image for adjusting the parameters of the depth estimation network. A training apparatus, and an electronic device applying the method are also disclosed.

Method for training depth estimation model, training apparatus, and electronic device applying the method

A lane line labeling method applied to an electronic device is provided. In the method, the electronic device acquires a target image corresponding to a target lane. Motion trajectory points of a target vehicle driving on the target lane are obtained. The electronic device determines projected pixel coordinates of the motion trajectory points on the target image, and determines target pixel coordinates corresponding to target lane lines on the target lane based on the projected pixel coordinates. Once target camera coordinates corresponding to the target pixel coordinates are obtained, the electronic device labels the target lane lines according to the target camera coordinates.

Lane line labeling method, electronic device and storage medium

A method for training an autoencoder implemented in an electronic device includes obtaining a stereoscopic image as the vehicle is in motion, the stereoscopic image includes a left image and a right image; generating a stereo disparity map according to the left image; generating a predicted right image according to the left image and the stereo disparity map; and calculating a first mean square error between the predicted right image and the right image.

Method for training autoencoder, electronic device, and storage medium

A method for processing images implemented in an electronic device includes obtaining images during moving of a vehicle; obtaining instance segmentation images by segmenting the images; obtaining a predicted disparity map by reconstructing the left images based on a pre-established autoencoder; generating a first error value of the autoencoder for the images according to the left image, the predicted disparity map, and the right image, generating a second error value of the autoencoder for the instance segmentation image according to the left image of instance segmentation, the predicted disparity map, and the right image of instance segmentation; establishing an autoencoder model by adjusting the autoencoder according to the first error value and the second error value; obtaining a test image as the vehicle is moving, and obtaining a target disparity map; and obtaining a depth image corresponding to the test image by converting the target disparity map.

Method for processing images, electronic device, and storage medium

A method and system for generating depth in monocular images acquires multiple sets of binocular images to build a dataset containing instance segmentation labels as to content; training an work using the dataset with instance segmentation labels to obtain a trained autoencoder network; acquiring monocular image, the monocular image is input into the trained autoencoder network to obtain a first disparity map and the first disparity map is converted to obtain depth image corresponding to the monocular image. The method combines binocular images with instance segmentation images as training data for training an autoencoder network, monocular images can simply be input into the autoencoder network to output the disparity map. Depth estimation for monocular images is achieved by converting the disparity map to a depth image corresponding to the monocular image. An electronic device and a non-transitory storage are also disclosed.

Method for generating depth in images, electronic device, and non-transitory storage medium

A method for obtaining depth images implemented in an electronic device includes obtaining a first image and a second image; obtaining a predicted depth map of the first image, and calculating a first error value of the predicted depth map; determining a first transformation matrix between the first image and the second image; obtaining an instance segmentation image and obtaining a first mask image and a second mask image; obtaining a target transformation matrix; converting the predicted depth map into a first point cloud image, converting the first point cloud image into a second point cloud image, and converting the second point cloud image into a third image; calculating a second error value between the second image and the third image; obtaining a target deep learning network model; and inputting at least one image into the target deep learning network model, and obtaining at least one depth image.

Method for obtaining depth images, electronic device, and storage medium

A driving route planning method applied to an electronic device is provided. The method includes acquiring an image when a vehicle is driving. Target route information is obtained by inputting the image into a target route planning model. Once a first embedding vector of the target route information is extracted, a driving route corresponding to a driving style is obtained by inputting the first embedding vector into a target driving style model.

Driving route planning method, electronic device and storage medium

A method for optimizing the conversion of a deep learning model to process other data, applied in a device, includes converting a first deep learning model to obtain a second deep learning model, obtaining a weighting arrangement of the two models according to their deep learning frameworks and performing a quantization on the two models. A similarity in weighting between the two models is analyzed to produce a weighting analysis based on the first and second weighting arrangement and the first and second model quantization result weighting. The two models are tested to establish a model performance analysis. One or more suggestions for optimization are obtained based on the weighting analysis and the model performance analysis, and are applied to optimize the second deep learning model, an optimized second deep learning model being employed to process the other data.

Method and device for optimizing deep learning model conversion, and storage medium

A method for training a depth estimation model is provided. The method includes obtaining a first left image and a first right image. A disparity map is obtained by inputting the first left image into a depth estimation model. A second right image is obtained by adding the first left image to the disparity map. The first left image is converted into a third right image. A mask image is obtained by performing a binarization processing on a pixel value of each of pixel points of the third right image. Once a loss value of the depth estimation model is obtained by calculating a mean square error of pixel values of all corresponding pixel points of the first right image, the second right image, and the mask image, a depth estimation model is iteratively trained according to the loss value.

Method for training depth estimation model, electronic device and readable storage medium

In a product defect detection method, a detection image of a product is obtained. A first preset number of detection blocks are cut out from the detection image. The detection blocks are input into an auto-encoder to obtain reconstructed blocks and a mean square error between each detection block and the corresponding reconstructed block is calculated, an association between the mean square error and the detection block being established. Whether the product carries defective is determined according to the mean square error of each detection block. The method improves accuracy of detecting defects of the product.

Method for detecting product defect in detection image, computing device, and storage medium

A method for detecting defects in products from images thereof and an electronic device applying the method inputs a defect image repair data set into an autoencoder to train the autoencoder, and generates a reconstructed image, calculates a reference error value between the sample image and the reconstructed image by a preset error function, and set a threshold value based on the reference error value. The electronic device inputs an image possibly revealing a defect into the autoencoder and generates the reconstructed image corresponding to the image to be detected, and uses the preset error function to calculate the reconstruction error between the image and the reconstructed image, thereby determining whether the image being analyzed does reveal defects. When the reconstruction error is greater than the threshold value, a determination is made that a defect is revealed.

Image defect detection method, electronic device using the same

A method for measuring a growth height of a plant, an electronic device, and a storage medium are provided. The method controls a camera device to obtain a color image and a depth image of a plant to be detected. The color image is detected by a detection model which is pre-trained, and a plurality of detection boxes which includes a plurality of plants to be detected is obtained. The color image and the depth image are aligned to create an alignment image. A plurality of target boxes is acquired from the alignment image, and depth values of the plurality of target boxes are determined. The quantity of the target boxes and a height of one or more plants to be detected are determined, no manual operations are required.

Method for measuring growth height of plant, electronic device, and storage medium

A method for neural network model encryption and decryption includes a first apparatus stores a neural network model and obtains hardware configuration information of the first apparatus, and obtains an encryption key accordingly; encrypts the neural network model by a predetermined encryption algorithm; a second apparatus obtains the encrypted neural network model from the first apparatus, transmits a decryption request to the first apparatus, obtains the hardware configuration information from the first apparatus, obtains a decryption key based on the hardware configuration information; and decrypts the encrypted neural network model by a predetermined decryption algorithm.

Method and apparatus for neural network model encryption and decryption

A method for generating defective image of products applied in an electronic device includes generating first input data according to flawless sample images and a first noise vector, using an autoencoder as a generator of a Generative Adversarial Network (GAN), inputting the first input data to the generator, and generating images for training in defects. The method further includes calculating a first loss value between the flawless sample images and the defect training images, inputting the defect training images into a discriminator of the GAN, and calculating a second loss value. The method further includes obtaining an optimized GAN and taking the optimized GAN as a defective image adversarial network, obtaining flawless testing images, inputting the flawless testing images and a second noise into a generator of the defective image adversarial network, and generating images of defects by processing the flawless testing images and the second noise.

Method for generating defective image, electronic device using the same

An image depth calculating device, which can calculate depth information of a binocular video with reduced computation, includes multiple modules. Modules receives a first frame information set that corresponds to a first frame time of the binocular video and establishes a first rhombic area that centers on a first pixel in a first viewing angle frame, to calculate a total pixel value of the first rhombic area. Modules further establishes a plurality of second rhombic areas that center on a plurality of second pixels of a pixel area in a second viewing angle frame to calculate a second total pixel value of each of the plurality of second rhombic areas. A depth calculating module compares the first total pixel value with each second total pixel value and calculates first frame depth information according to the result. An image depth calculating method is also provided.

Growing community of inventors

Taipei, Taiwan

Jung-Hao Yang