大图裁剪为小图输入网络

bug：

开始时写成了
`small_imgs = torch.zeros(images.shape[0],images.shape[1],opt['network']['H'],opt['network']['H'],num).to('cuda')`
图像切花了

```go
reshape的时候要把多的一维放到batch旁边，torch默认最后两维是H、W‘’

#
def crop_imgs(images, opt):
    scal = int(opt['datasets']['H']/opt['network']['H'])
    num = scal*scal
    small_imgs = torch.zeros(images.shape[0],num,images.shape[1],opt['network']['H'],opt['network']['H']).to('cuda')
    count = 0
    for i in range(scal):
        for j in range(scal):
            small_imgs[:,count,:,:,:] = images[:, :, i*opt['network']['H']:(i+1)*opt['network']['H'], j*opt['network']['H']:(j+1)*opt['network']['H']] 
            count+=1

    small_imgs = torch.reshape(small_imgs, [small_imgs.shape[0]*small_imgs.shape[1], images.shape[1],opt['network']['H'],opt['network']['W']])
    return small_imgs