48 results about "Darknet" patented technology

The invention discloses a traffic sign detection algorithm based on an improved YOLO v3 algorithm. According to the method, a feature extraction network capable of maintaining high-resolution representation is designed to replace DarkNet-53 in an original YOLO v3 algorithm, so that the detection precision of a small-size target traffic sign is improved, and the parameter quantity of the algorithmis reduced; attention of the detection algorithm to small and medium-sized targets is increased by fusing the feature maps participating in prediction; and optimizing the loss function by using a GIOUalgorithm and a local loss algorithm. According to the method, the detection accuracy of the small-size traffic sign is improved, and the traffic sign can be rapidly and accurately detected and identified on a complex traffic road.

The invention discloses a pedestrian detection method based on deep learning multi-network soft fusion, and relates to the technical field of image processing, target detection and deep learning. Themethod comprises the following steps: S1, inputting a to-be-processed image; S2, inputting the to-be-processed image into a YOLO v3 pedestrian candidate region generator taking Darknet-53 as a basic network, and generating a pedestrian candidate region; S3, inputting the to-be-processed image into a front-end prediction module, and outputting C feature maps; s4, inputting the C feature maps into asemantic segmentation system, and outputting C feature maps containing context information; s5, fusing a result of the semantic segmentation system with a pedestrian candidate result generated by a pedestrian candidate region generator; s6: outputting a detection image. According to the method, the pedestrian candidate region generator and the semantic segmentation system are subjected to parallel soft fusion, pedestrians in various challenge scenes are efficiently detected, and meanwhile, the small target detection capability is improved.

The invention discloses a darknet clue detection method based on heterogeneous graph attention neural network. Extract and construct a dynamic heterogeneous information network; step 3, perform embedding processing on the nodes in the constructed heterogeneous information network, and obtain feature vectors of each node; step 4, learn the graph structure of the heterogeneous information network; step 5 . According to the result of learning the graph structure of the heterogeneous information network, the nodes in the heterogeneous information network are classified into clue categories, so as to complete the clue detection of darknet information. The invention utilizes the external knowledge base as the support, and adopts two sets of methods to learn the graph structure of the constructed heterogeneous information network, and has good clue detection effect.

The invention discloses an image classification method integrating DarkNet and CapsuleNet models, which belongs to the technical field of image processing and solves the problem of poor classification effect due to unbalanced data. The technical scheme includes the following steps: constructing a DarkNet-Capsule network fusion classification model, realizing Define the loss function of the fusion classification model, input the image to be classified in the fusion classification model, use DarkNet for forward training, and extract the feature map of the target image; further process the feature map of the target image, and complete the error back propagation update through loss. parameters of the entire network; after multiple rounds of iterative learning, the fusion classification model is used to complete the image classification; in the field of image classification, the invention can further improve the classification accuracy when the data is unbalanced, and also lays a solid foundation for the research of machine vision. a more solid foundation.

The invention discloses a target detection method and system based on multi-scale feature map reconstruction and knowledge distillation. The method first uses the backbone network Darknet-53 to extract features, and the deep features generate multi-scale features through upsampling and shallow feature tensor splicing. Feature map; then use the feature re-calibration strategy to automatically obtain the weight of each channel in the feature map, promote useful features and suppress useless features according to the weight, and then use the residual module to fuse the semantic information of the top-level features and the details of the underlying features; Then, the γ coefficient of the batch normalization layer in the backbone network is introduced into the pruning objective function for training, and the channel where the γ coefficient below the threshold is located is removed from the model according to the pruning threshold; finally, the trained YOLOv3 benchmark model is used as the teacher. network, the pruned model is used as a student network for knowledge distillation. The invention improves the accuracy of detecting objects of different sizes in a large range, reduces the calculation amount of the model, and improves the model detection speed.