A static image crowd counting method based on joint learning

Abstract

The invention discloses a static image crowd counting method based on joint learning, includes the following steps: pre-training stage: using 50-layer residual network to train the ImageNet 2012 classification data set to obtain the parameter initialization classifier network, classifying the image blocks into three categories by a Softmax, respectively corresponding to three regressors; In the training phase, every image block of the training data set is input to three regressors respectively, and the same image block will get different counting results. The regressor with the smallest counting error is used as a classification tag to mark the image block, and the three kinds of image blocks are used to fine-tune the regressors respectively. Classifier training phase: random sampling andensure that the number of labels of each category is consistent; Joint training phase: continuous iterative training of classifiers and regressors. The invention can carry out counting and density estimation in crowd dense scene, has certain scale self-adaptability, and improves counting precision and model generalization ability.

Description

technical field [0001] The invention relates to a static image crowd counting method based on joint learning. Background technique [0002] Crowd counting has important social significance and market application prospects. Making full use of people counting information can provide effective guidance for safety warnings in crowded shopping malls, stations, squares and other public places, and can also bring economic benefits, such as improving service quality , Analyze customer behavior, place advertisements and optimize resource allocation, etc. In addition, crowd counting methods can also be extended to other areas, such as counting cells or bacteria from microscopic images, estimating animal populations in wildlife sanctuaries, and estimating the number of vehicles in traffic hubs and traffic jams. [0003] Static image crowd counting generally includes two tasks: crowd counting and density estimation. The purpose of crowd counting is to count the number of people in a s...

AI Technical Summary

Problems solved by technology

[0012] 3. Overfitting phenomenon

The application of deep learning methods to crowd counting tasks has achieved great success, but deep learning has certain requirements for the amount of data in the data set. Insufficient data will lead to network overfitting problems, making the generalization ability of the model poor

However, due to the difficulty of manual labeling, there is insufficient data in the densely populated scene dataset. For example, the UCF_50_CC dataset only contains 50 images.

Images