Created by: liuzengzhen1

1.根据文档下载pascalvoc数据，以及预训练模型 2.调试相应参数执行train.py

----------- Configuration Arguments ----------- ap_version: 11point batch_size: 8 data_dir: D:/untitled/pratice/ssd/data/pascalvoc dataset: pascalvoc enable_ce: False epoc_num: 1000 image_shape: 3,300,300 learning_rate: 0.001 mean_BGR: 127.5,127.5,127.5 model_save_dir: D:\untitled\pratice\ssd\model parallel: 0 pretrained_model: D:\untitled\pratice\ssd\pretrained\mobilenet_v1_imagenet use_gpu: 0

3.执行结果： Batch 10, map [0.00233454] Batch 20, map [0.00179878] Batch 30, map [0.00124445] Batch 40, map [0.00154742] Batch 50, map [0.00117008] Batch 60, map [0.00101512] Batch 70, map [0.0009321] Batch 80, map [0.00084877] Batch 90, map [0.00079377] Batch 100, map [0.00078082] Batch 110, map [0.00077564] Batch 120, map [0.00079855] Batch 130, map [0.00077792] Batch 140, map [0.00077779] Batch 150, map [0.00077369] Batch 160, map [0.00073959] Batch 170, map [0.00072543] Batch 180, map [0.00071779] Batch 190, map [0.00069148] Batch 200, map [0.00067471] Batch 210, map [0.00065191] Batch 220, map [0.00062474] Batch 230, map [0.00062316] Batch 240, map [0.00062389]

请问执行准确率这么低是什么原因？，是我操作传入参数有问题么？另外这个模型提供预训练模型，我想看下实际的准确率情况能否直接infer.py，将model_dir更改为预训练模型路径？我尝试了下报错，是不能这样验证么？