Pytorch classification model plotting confusion matrix and visualization details

# First define a null confusion matrix of classifications*classifications
 conf_matrix = (Emotion_kinds, Emotion_kinds)
 # Use torch.no_grad() to significantly reduce the GPU footprint of test cases
    with torch.no_grad():
        for step, (imgs, targets) in enumerate(test_loader):
            # imgs:     ([50, 3, 200, 200])   
            # targets: ([50, 1]), one more dimension, so we're going to get rid of it
            targets = ()  # [50,1] ----->  [50]

            # Convert variables to gpu
            targets = ()
            imgs = ()
            # print(step,,(),,())
            
            out = model(imgs)
            #Record Confusion Matrix Parameters
            conf_matrix = confusion_matrix(out, targets, conf_matrix)
            conf_matrix=conf_matrix.cpu()

def confusion_matrix(preds, labels, conf_matrix):
    preds = (preds, 1)
    for p, t in zip(preds, labels):
        conf_matrix[p, t] += 1
    return conf_matrix

conf_matrix=(conf_matrix.cpu())# Transfer obfuscation matrix from gpu to cpu to np
corrects=conf_matrix.diagonal(offset=0)# of correct identifications per classification for extracted diagonals
per_kinds=conf_matrix.sum(axis=1)# of total test strips extracted for each categorized data

 print("Total number of elements in the confusion matrix：{0},Total number of test sets:{1}".format(int((conf_matrix)),test_num))
 print(conf_matrix)

 # Get recognition accuracy for each Emotion
 print("Total number of each emotion:",per_kinds)
 print("Number of correct predictions for each emotion:",corrects)
 print("The recognition accuracy for each emotion is：{0}".format([rate*100 for rate in corrects/per_kinds]))

# Drawing the confusion matrix
Emotion=8#This value is a specific number of categories, you can change it yourself
labels = ['neutral', 'calm', 'happy', 'sad', 'angry', 'fearful', 'disgust', 'surprised']# Labels for each category

# Display data
(conf_matrix, cmap=)

# Quantitative/probabilistic information labeled in the graphs
thresh = conf_matrix.max() / 2	# Numeric color threshold, if the value exceeds this, the color is deepened.
for x in range(Emotion_kinds):
    for y in range(Emotion_kinds):
        # Note that matrix[y, x] is not matrix[x, y] here
        info = int(conf_matrix[y, x])
        (x, y, info,
                 verticalalignment='center',
                 horizontalalignment='center',
                 color="white" if info > thresh else "black")
                 
plt.tight_layout()# Ensure that the diagrams don't overlap
(range(Emotion_kinds), labels)
(range(Emotion_kinds), labels,rotation=45)# X-axis font is tilted 45°
()
()