调整了tobacco,yolo模型和YOLO的置信度

This commit is contained in:
FEIJINTI 2022-08-27 14:58:29 +08:00
parent 12870b3880
commit 210c7354cf
2 changed files with 19 additions and 12 deletions

View File

@ -54,14 +54,14 @@
"dataset_file = \"data/dataset/dataset_2022-07-20_10-04.mat\"\n",
"\n",
"# color_dict = {(0, 0, 255): \"yangeng\", (255, 0, 0): 'beijing',(0, 255, 0): \"zibian\"} # 颜色对应的类别\n",
"# color_dict = {(0, 0, 255): \"yangeng\"}\n",
"color_dict = {(255, 0, 0): 'beijing'}\n",
"color_dict = {(0, 0, 255): \"yangeng\"}\n",
"# color_dict = {(255, 0, 0): 'beijing'}\n",
"# color_dict = {(0, 255, 0): \"zibian\"}\n",
"label_index = {\"yangeng\": 1, \"beijing\": 0, \"zibian\":2} # 类别对应的序号\n",
"show_samples = False # 是否展示样本\n",
"\n",
"# 定义一些训练量\n",
"threshold = 2 # 正样本周围多大范围内的还算是正样本\n",
"threshold = 3 # 正样本周围多大范围内的还算是正样本\n",
"node_num = 20 # 如果使用ELM作为分类器物有多少的节点\n",
"negative_sample_num = None # None或者一个数字对应生成的负样本数量"
],
@ -169,7 +169,7 @@
"name": "stderr",
"output_type": "stream",
"text": [
"147099it [31:23, 78.09it/s] \n"
"23961it [00:46, 519.87it/s] "
]
},
{
@ -178,12 +178,19 @@
"text": [
" precision recall f1-score support\n",
"\n",
" 0 1.00 1.00 1.00 44129\n",
" 1 1.00 1.00 1.00 36775\n",
" 0 1.00 1.00 1.00 7188\n",
" 1 1.00 1.00 1.00 5991\n",
"\n",
" accuracy 1.00 80904\n",
" macro avg 1.00 1.00 1.00 80904\n",
"weighted avg 1.00 1.00 1.00 80904\n",
" accuracy 1.00 13179\n",
" macro avg 1.00 1.00 1.00 13179\n",
"weighted avg 1.00 1.00 1.00 13179\n",
"\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"\n"
]
}

View File

@ -25,13 +25,13 @@ class Config:
spec_size_threshold = 3
# rgb模型参数
rgb_tobacco_model_path = r"weights/tobacco_dt_2022-08-05_10-38.model"
rgb_tobacco_model_path = r"weights/tobacco_dt_2022-08-27_14-43.model"
rgb_background_model_path = r"weights/background_dt_2022-08-22_22-15.model"
threshold_low, threshold_high = 10, 230
threshold_s = 190 # 饱和度的最高允许值
rgb_size_threshold = 4 # rgb的尺寸限制
ai_path = 'weights/best0825.pt'
ai_conf_threshold = 0.5
ai_path = 'weights/best0827.pt'
ai_conf_threshold = 0.6
# mask parameter
target_size = (1024, 1024) # (Width, Height) of mask