导盲赛道思路分享

## 解压文件夹!tar -zxvf data/data137625/WisdomGuide.tar.gz## 安装所需环境!pip install pycocotools

from pycocotools.coco import COCO# 查看train数据分布annFile='WisdomGuide/annotations/instance_train.json'coco=COCO(annFile)
training_data= {}
cats = coco.loadCats(coco.getCatIds()) 
cat_nms=[cat['name'] for cat in cats] 
print('-'*10,"training data",'-'*10)for cat_name in cat_nms:
    catId = coco.getCatIds(catNms=[cat_name])
    imgId = coco.getImgIds(catIds=catId)
    annId = coco.getAnnIds(imgIds=imgId, catIds=catId, iscrowd=None)
    training_data[cat_name] = len(imgId), len(annId)    print("{:<15} {:<6d}     {:<10d}".format(cat_name, len(imgId), len(annId)))print(training_data)# 查看val数据分布annFile='WisdomGuide/annotations/instance_val.json'coco=COCO(annFile)

cats = coco.loadCats(coco.getCatIds()) 
val_data = {}
cat_nms=[cat['name'] for cat in cats] 
print('-'*10,"valuation data",'-'*10)for cat_name in cat_nms:
    catId = coco.getCatIds(catNms=[cat_name])
    imgId = coco.getImgIds(catIds=catId)
    annId = coco.getAnnIds(imgIds=imgId, catIds=catId, iscrowd=None)
    val_data[cat_name] = (len(imgId), len(annId))            
    print("{:<15} {:<6d}     {:<10d}".format(cat_name, len(imgId), len(annId)))

import matplotlib.pyplot as plt
name = [x for x in training_data.keys()]
train_imgid = [training_data[name[x]][0] for x in range(len(name))]
train_annid = [training_data[name[x]][1] for x in range(len(name))]
val_imgid = [val_data[name[x]][0] for x in range(len(name))]
val_annid = [val_data[name[x]][1] for x in range(len(name))]

plt.figure(figsize=(6,6))#将画布设定为正方形，则绘制的饼图是正圆values = train_imgid
label = name
explode = [0.01,0.01,0.01,0.01,0.01]
patches,l_text,p_text = plt.pie(values,explode=explode,labels=label,autopct='%1.2f%%')
plt.suptitle('train_imgid',fontsize=16,y=0.93)
plt.legend(bbox_to_anchor=(-0.04, 1),borderaxespad=0,frameon=False)
plt.show()

plt.figure(figsize=(6,6))
values = train_annid
label = name
explode = [0.01,0.01,0.01,0.01,0.01]
patches,l_text,p_text = plt.pie(values,explode=explode,labels=label,autopct='%1.2f%%')
plt.suptitle('train_annid',fontsize=16,y=0.93)
plt.legend(bbox_to_anchor=(-0.04, 1),borderaxespad=0,frameon=False)
plt.show()

plt.figure(figsize=(6,6))
values = val_imgid
label = name
explode = [0.01,0.01,0.01,0.01,0.01]
patches,l_text,p_text = plt.pie(values,explode=explode,labels=label,autopct='%1.2f%%')
plt.suptitle('val_imgid',fontsize=16,y=0.93)
plt.legend(bbox_to_anchor=(-0.04, 1),borderaxespad=0,frameon=False)
plt.show()

plt.figure(figsize=(6,6))
values = val_annid
label = name
explode = [0.01,0.01,0.01,0.01,0.01]
patches,l_text,p_text = plt.pie(values,explode=explode,labels=label,autopct='%1.2f%%')
plt.suptitle('val_annid',fontsize=16,y=0.93)
plt.legend(bbox_to_anchor=(-0.04, 1),borderaxespad=0,frameon=False)
plt.show()

import paddle 
import paddle.nn as nnimport time
paddle.device.set_device("cpu")class SPP(nn.Layer):
    def __init__(self):
        super().__init__()
        self.maxpool1 = nn.MaxPool2D(5,1,padding=2)
        self.maxpool2 = nn.MaxPool2D(9,1,padding=4)
        self.maxpool3 = nn.MaxPool2D(13,1,padding=6)    def forward(self, x):
        p = x
        o1 = self.maxpool1(x)
        o2 = self.maxpool2(x)
        o3 = self.maxpool3(x)        return paddle.concat([x,o1,o2,o3],axis=1)class SPPF(nn.Layer):
    def __init__(self):
        super().__init__()
        self.maxpool = nn.MaxPool2D(5, 1, padding=2)    def forward(self, x):
        o1 = self.maxpool(x)
        o2 = self.maxpool(o1)
        o3 = self.maxpool(o2)        return paddle.concat([x, o1, o2, o3], axis=1)def main():
    input_tensor = paddle.rand((8, 32, 16, 16))
    spp = SPP()
    sppf = SPPF()
    output1 = spp(input_tensor)
    output2 = sppf(input_tensor)    # print(paddle.equal(output1, output2))

    t_start = time.time()    for _ in range(100):
        spp(input_tensor)    print(f"spp time: {time.time() - t_start}")

    t_start = time.time()    for _ in range(100):
        sppf(input_tensor)    print(f"sppf time: {time.time() - t_start}")if __name__ == '__main__':
    main()