Pretrained Densenet Problem

1

Please some one help me.

I have written the following code.

import torch
import torchvision.models as models
import numpy as np
import matplotlib.pyplot as plt
import torchvision.transforms as transforms
import cv2

image=cv2.imread(‘Dog.jpg’)

image is numpy array

cv2.imshow(‘Display’,image);cv2.waitKey(10000);cv2.destroyAllWindows()
image1 = transforms.ToTensor()(image)

image1 is Tensor

image2=transforms.ToPILImage()(image1)

image2 is PIL image

image3=transforms.Resize((224,224))(image2)
#image3 is Resized Image
image4 = transforms.ToTensor()(image3)

image4 is Tensor

image4=transforms.Normalize(mean=[0.485, 0.456, 0.406],std=[0.229, 0.224, 0.225])(image4)
image4=torch.unsqueeze(image4,0)
densenet=models.densenet161(pretrained=True)
output=densenet(image4)
predictions=output.max(1)

##########################################################
Dog image is as shown…

Dog
Dog1280×620 72.4 KB

I got the output as:
predictions : (tensor([ 2.0526]), tensor([ 463]))

From standard Imagenet labels ,463 corresponds to bucket or pail.

Why I am not getting coorect label???

2

After loading your model, you should set it into evaluation mode with:

densenet.eval()

I’ve tested your code and the result for this image is class 207, which is golden retriever.

1 Like
3