8 results about "Automatic image annotation" patented technology

The invention discloses a positive sample and large model weak supervision-based rigid object automatic image labeling method. The method comprises the following steps of: 1, acquiring a rigid object image; step 2, segmenting each rigid object image by using an SAM model; step 3, firstly screening Masks which are not in a set range through an area and a gravity center, and then extracting a positive sample in a manual interaction mode; step 4, uniformly segmenting each positive sample Mask into a plurality of patch patches with the pixel size of r * r; 5, constructing an anomaly detection model, wherein the input of the anomaly detection model is a patch with the pixel size of r * r, and the output of the anomaly detection model is dissimilarity with a positive sample; step 6, performing anomaly detection model training by using all the patch patches after the Mask segmentation of all the positive samples; and 7, performing image annotation in combination with the SAM and the anomaly detection model. The method does not depend on excessive manual intervention, automatic labeling of the whole data set can be achieved only by screening a very small number of positive samples in the initial stage, and the labeling efficiency is improved.

This application provides an automatic image annotation method and apparatus. The method includes acquiring an image to be annotated; identifying a first position of a target feature in the image by performing target feature detection on the image to be annotated; determining a bounding box corresponding to the target feature at the first position according to a preset annotation rule; determining an annotation strategy based on the distribution type of the target feature in the image to be annotated; selecting a target bounding box from the bounding boxes determined according to the preset annotation rule according to the annotation strategy; and annotating the target feature in the image to be annotated using the target bounding box. This method can accurately and automatically annotate target features in an image, with low cost, high speed, and high accuracy.

The invention provides a visual pollution AI identification system and device based on YOLOv8 and OpenCV, a YOLOv8 model is used for preliminary training to form a reference model, the reference model is reversely used for automatic image annotation, and a high-quality closed-loop data system is constructed through continuous iterative training in combination with manual verification; a self-developed image gray, binarization processing and pixel proportion statistical analysis algorithm is used to assist in identifying different types of flue gas forms; more shape information is extracted through polygonal region labeling and morphological analysis, so that the model can adapt to complex backgrounds and unseen scenes; and a lightweight model and integrated hardware equipment are selected, so that special adaptive design of industrial hardware and an inference server is realized, and modular rapid deployment can be carried out.

This application discloses an automatic image annotation method, apparatus, device, and storage medium. The method includes: annotating an original screenshot and saving the annotation information and image resolution of the original screenshot, the annotation information including annotation coordinate positions; recognizing the text information annotated in the original screenshot, the text information including text content and text coordinate positions; obtaining the image resolution of a new screenshot and recognizing the text content of the new screenshot to obtain the text information of the new screenshot; calculating a scaling ratio based on the image resolutions of the original and new screenshots, and obtaining the number of occurrences of the text content corresponding to the annotation coordinate positions in the original screenshot in the new screenshot by comparing the text content in the original screenshot with the text content in the new screenshot; obtaining the annotation information of the new screenshot based on the number of occurrences and the scaling ratio, and annotating the new screenshot based on the annotation information of the new screenshot. This improves the technical problem of low annotation efficiency in existing technologies that rely on manual image annotation.

The invention discloses an automatic image annotation method and device, computer equipment and a storage medium, belongs to the technical field of artificial intelligence, and is applied to processing of vehicle insurance loss assessment images. According to the method, irrelevant information is removed by screening and classifying the original image data. And then, key appearance parts of the vehicle are accurately positioned by using a deep learning segmentation and recognition technology, and specific damaged parts are matched and positioned in combination with the three-dimensional model, so that part-level refined analysis is realized. On the basis, a damage area is extracted through local screenshot, image matching is conducted through a preset standard part damage library, the damage category and the damage degree are automatically recognized, and structured damage information is generated. And finally, integrally labeling and outputting the damage position, category and degree label and the target image set to form a high-quality labeled image set. According to the method and the device, efficient and unified automatic labeling can be carried out on mass traffic accident pictures in a short time, and the picture labeling efficiency and consistency are remarkably improved.

The present invention discloses a method and system for automatic image annotation, characterized by comprising the following steps: acquiring images and placing them into a pool of images to be classified, screening and classifying the images; introducing new image categories and reference images based on the needs of the business model; calculating feature vectors of the reference images and the images to be classified; calculating the similarity between the images to be classified and the reference images, and automatically assigning the images to be classified to the category containing the reference image with the highest similarity; batch reviewing the results of the automatic machine classification to obtain a confirmed image dataset; and training a new model based on the confirmed image dataset. The present invention can efficiently and accurately complete image annotation, significantly reducing labor costs and accelerating the implementation of artificial intelligence applications.

The invention relates to the field of computer vision and machine learning, in particular to an image annotation method based on region segmentation and FAISS optimization lookup, and the method comprises the steps: carrying out the adaptive segmentation of a to-be-annotated image through OpenCV, carrying out the scanning of connected region marks twice, decomposing the image into a plurality of semantic independent regions, and carrying out the lookup of each region; constructing a two-stage FAISS index mechanism, introducing a ConvNeXt classification result as a prior constraint, and preferentially carrying out one-time retrieval in a same-class sample range; a dynamic clustering strategy is adopted, the number of clustering centers is adjusted in a self-adaptive mode according to the data scale, and a long-tail large cluster is dynamically split; and taking the distance between the similar image and the query image as a weight, carrying out weighted voting on the retrieved labels, fusing ConvNeXt classification confidence to dynamically adjust the retrieval weight, obtaining high-confidence labels of each region, and combining to form a whole image labeling result. According to the method, a region-level independent retrieval and dynamic clustering scattering mechanism is introduced, so that multi-target labeling is more sensitive and clearer; and a ConvNeXt classification result is utilized to limit a search space, so that cross-class missing check is avoided, and efficient and accurate automatic image annotation is realized.

This invention discloses an automatic image annotation method based on concept embedding search and active learning. The method includes the following steps: constructing an initial labeled sample set, and introducing a learnable embedding E based on the original concept embedding TEnc(T), forming a modified concept embedding TEnc(T)+E; performing M2C optimization on the learnable embedding E to obtain an updated E. t ; Utilizing the updated E t The concept embedding TEnc(T) is superimposed onto the original concept embedding to drive the SAM3 model to automatically generate concept-driven prediction segmentation masks for the remaining samples; a final mixed uncertainty score is calculated to comprehensively evaluate the difficulty of the samples; the prediction results are sorted, and the top-ranked samples are manually checked and corrected; this is fed back to step S2), driving the corrected concept embedding TEnc(T)+E to enter the next round of continuous search and update. This invention greatly reduces the cost of manual annotation while significantly improving the accuracy and robustness of image segmentation.