31results about How to "Enrich training data" patented technology

The present invention proposes a multi-mode security monitoring method based on deep learning image processing, which divides the scenes that need to be monitored according to the degree of importance, and regularly adjusts them according to the follow-up machine learning results, so as to realize the application of limited monitoring resources in the most needed monitoring At the same time, focus on the content that needs to be monitored. By setting the dangerous ID and the sub-dangerous ID, the people who are most likely to be monitored will be screened out. Focus on the people who need to be monitored the most. At the same time, by collecting the actual scenes that need to be monitored and pasting a large amount of external training data, the data of the pre-training set that meets the scenes to be monitored is enriched, thereby improving the accuracy of the final recognition. At the same time, through the training results, it is also possible to adjust and supervise the areas most likely to send abnormal behavior. Realize video online monitoring with lower equipment cost and overhead to achieve maximum efficiency.

An image simulation generation method provided by the present invention, which determines the threshold of the degree of simulation based on the preliminary analysis of the initial image, determines the predetermined pixel value of the translation of the initial image in at least two directions based on the threshold of the degree of simulation, and translates The final images are weighted and averaged to output the desired simulated image. The above-mentioned method can solve the technical problem that the existing image simulation technology cannot effectively simulate image aberrations caused by lenses, optics, photosensitive elements, and the like. The image simulation generation system and its electronic equipment have the same technical effect as the method. The deep learning algorithm training method provided by the present invention can also solve the problem that the existing data amplification technology cannot simulate the generated image into a real sampling image well, so as to obtain a better training effect.

The present invention provides a training method for a recommendation model, comprising: initializing parameters of the training model, wherein the training model includes a first self-encoder, a first item encoder, a first user encoder, a first decoder, a second The self-encoder, the second item encoder, the second user encoder and the second decoder; obtain training data, wherein the training data includes the first initial score vector of the source domain, the first user content vector, and the first item content vector and the second initial scoring vector, the second user content vector, and the second item content vector of the target domain; the training model is trained according to the training data, and the second user encoder, the second item encoder and the second decoder are used as recommendation Model. The technical scheme of the present invention effectively solves the data sparse problem and the cold start problem.

The invention discloses a human activity gesture recognition method based on a multi-level end-to-end neural network, comprising the following steps: collecting labeled data of human activity gestures, cutting the labeled data by using a sliding window, and obtaining several segments of equidistant A first labeled data window, processing the first labeled data window using a gait-based data augmentation algorithm to obtain segments of a second labeled data window; using the first labeled data window and the second labeled data window Windowed pair multi-layer end-to-end neural network training. Collect the raw data of any human activity posture, and perform sliding window cutting on the raw data to obtain several consecutive motion data windows to be recognized; import the motion data windows to be recognized into the trained multi-level terminal in sequence Peer-to-peer neural network to identify the type of human activity gestures. The invention has the advantages of high recognition accuracy, low calculation complexity, low power consumption, etc., and has broad market prospects in the fields of medical technology, behavior supervision and the like.

The present invention relates to the technical field of data processing, and specifically discloses a method, system, electronic device and storage medium for predicting properties of luminescent materials, including the following steps: input sample data of luminescent materials into the first deep learning model for training, and obtain the second Deep learning model; obtain the first image data of the simplified chemical structure of the luminescent material to be predicted; enhance the first image data to obtain second image data; input the second image data into the second deep learning model , to predict the luminescent properties of luminescent materials. The technical solution provided by the present invention can automatically predict the luminescent properties of luminescent materials by adopting the artificial intelligence method of deep learning, thereby greatly reducing the cost of manually verifying the properties of luminescent materials, thereby accelerating the process of research and development of luminescent materials.

The invention provides a textile quality detection method and system based on image recognition. The method includes: obtaining a first inspection environment; obtaining a first product to be inspected, placing it in the first inspection environment, and obtaining a first image; Color feature recognition, obtaining the first set of abnormal color feature points, and collecting the enlarged image of a specific point corresponding to the position of the first product to be inspected to obtain the first set of abnormal images; obtaining the first set of abnormal images according to the first set of abnormal images The first detection result; extracting the fiber boundary from the first image to obtain the first extraction result, and obtaining the fabric uniformity detection result; obtaining the quality of the first product to be inspected through the first detection result and the fabric uniformity detection result Test results. Due to the fact that the prior art mainly uses chemical detection, there is a technical problem of inapplicability when faced with factories with a large sample volume.