Python進行圖檔t-SNE降維可視化

文章目錄

• 問題描述
• 解決方案
• IPython代碼
• 參考文獻

問題描述

Python進行圖檔t-SNE降維可視化

解決方案

下載下傳資料集 plant-seedlings-classification 後解壓，把 train.zip 放在根目錄下解壓

IPython代碼

%matplotlib inline

import os
import cv2
import matplotlib
import numpy as np
from glob import glob
import matplotlib.cm as cm
import matplotlib.pyplot as plt
from sklearn.manifold import TSNE
from sklearn.decomposition import PCA
from sklearn.preprocessing import StandardScaler
from matplotlib.offsetbox import OffsetImage, AnnotationBbox
           

BASE_DATA_FOLDER = "./"
TRAIN_DATA_FOLDER = os.path.join(BASE_DATA_FOLDER, "train")
           

def create_mask_for_plant(image):
    image_hsv = cv2.cvtColor(image, cv2.COLOR_BGR2HSV)

    sensitivity = 35
    lower_hsv = np.array([60 - sensitivity, 100, 50])
    upper_hsv = np.array([60 + sensitivity, 255, 255])

    mask = cv2.inRange(image_hsv, lower_hsv, upper_hsv)
    kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (11,11))
    mask = cv2.morphologyEx(mask, cv2.MORPH_CLOSE, kernel)
    
    return mask

def segment_plant(image):
    mask = create_mask_for_plant(image)
    output = cv2.bitwise_and(image, image, mask = mask)
    return output

def visualize_scatter(data_2d, label_ids, figsize=(20,20)):
    plt.figure(figsize=figsize)
    plt.grid()
    
    nb_classes = len(np.unique(label_ids))
    
    for label_id in np.unique(label_ids):
        plt.scatter(data_2d[np.where(label_ids == label_id), 0],
                    data_2d[np.where(label_ids == label_id), 1],
                    marker='o',
                    color= plt.cm.Set1(label_id / float(nb_classes)),
                    linewidth='1',
                    alpha=0.8,
                    label=id_to_label_dict[label_id])
    plt.legend(loc='best')
           

images = []
labels = []

for class_folder_name in os.listdir(TRAIN_DATA_FOLDER):
    class_folder_path = os.path.join(TRAIN_DATA_FOLDER, class_folder_name)
    for image_path in glob(os.path.join(class_folder_path, "*.png")):
        image = cv2.imread(image_path, cv2.IMREAD_COLOR)
        
        image = cv2.resize(image, (150, 150))
        image = segment_plant(image)
        image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
        image = cv2.resize(image, (45,45))
        
        image = image.flatten()
        
        images.append(image)
        labels.append(class_folder_name)
        
images = np.array(images)
labels = np.array(labels)
           

指定圖檔格式為.png

等待運作完畢

label_to_id_dict = {v:i for i,v in enumerate(np.unique(labels))}
id_to_label_dict = {v: k for k, v in label_to_id_dict.items()}

label_ids = np.array([label_to_id_dict[x] for x in labels])
           

id_to_label_dict
           

{0: 'Black-grass',
 1: 'Charlock',
 2: 'Cleavers',
 3: 'Common Chickweed',
 4: 'Common wheat',
 5: 'Fat Hen',
 6: 'Loose Silky-bent',
 7: 'Maize',
 8: 'Scentless Mayweed',
 9: 'Shepherds Purse',
 10: 'Small-flowered Cranesbill',
 11: 'Sugar beet'}
           

images_scaled.shape
           

(4750, 2025)
           

label_ids.shape
           

(2435,)
           

pca = PCA(n_components=180)
pca_result = pca.fit_transform(images_scaled)
           

pca_result.shape
           

(4750, 180)
           

tsne = TSNE(n_components=2, perplexity=40.0)
tsne_result = tsne.fit_transform(pca_result)
tsne_result_scaled = StandardScaler().fit_transform(tsne_result)
visualize_scatter(tsne_result_scaled, label_ids)
           

等待運作完畢

其他圖請自行查閱參考文獻：

3D圖，gif動圖看原文獻

參考文獻

1. Plants PCA & t-SNE | Kaggle
2. 從SNE到t-SNE再到LargeVis