-
图 1 网络整体结构
Figure 1.
-
图 2 注意力模型网络结构
Figure 2.
-
图 3 分级网络中每一级的识别结果
Figure 3.
-
图 4 数据集部分属性间相关性及其共现概率
Figure 4.
-
图 5 网络参数及改进效果对比
Figure 5.
-
属性类(G) 属性 数量(k) Gender male, female 2 Age child, teenager, adult, old 4 Hair length long, short 2 Length of lower-body clothing long, short 2 Type of lower-body clothing pants, dress 2 Wearing hat yes, no 2 Carrying bag yes, no 2 Carrying backpack yes, no 2 Carrying handbag yes, no 2 Color of upper-body clothing black, white, red, purple,yellow, gray, blue, green 8 Color of lower-body clothing black, white, pink, purple,yellow, gray, blue, green, brown 9 表 1 Market1501数据集中的属性类别
-
行人属性 gender age hair L.slv L.low S.cloth B.pack H.bag bag hat C.up C.low mean 基础网络 82.18 85.32 80.12 92.48 71.58 85.67 79.57 81.54 79.66 70.56 91.23 87.81 82.31 本文算法 90.27 88.15 91.54 93.55 87.25 90.48 89.77 87.65 84.67 87.39 92.44 93.48 89.72 表 2 Market1501数据集各属性识别准确率(%)
-
行人属性 gender hat boots L.up B.pack H.bag bag C.shoes C.up C.low mean 基础网络 82.47 75.48 76.14 73.58 71.58 69.42 78.31 68.54 62.17 51.24 70.89 本文算法 83.59 87.24 84.56 76.33 77.11 75.32 83.78 72.19 74.88 62.18 77.72 表 3 DukeMTMC数据集各属性识别准确率(%)
-
表 4 Market1501数据集行人再识别结果(%)
-
表 5 DukeMTMC数据集行人再识别结果(%)
图共
5 个 表共
5 个