bigdataman

README.md

@@ -1,3 +1,57 @@
 ---
 license: mit
 ---
+```
+!pip list
+!pip install matplotlib
+!pip install --no-build-isolation scikit-learn
+!pip install numpy scipy cython
+!pip install mxnet
+!pip install opencv-python
+!pip install numpy==1.23.5
+!pip install mxnet
+import mxnet as mx
+from mxnet import recordio
+import matplotlib.pyplot as plt
+import cv2
+import os
+
+path_imgidx = 'train.idx' # path to train.rec
+path_imgrec = 'train.rec' # path to train.idx
+
+imgrec = recordio.MXIndexedRecordIO(path_imgidx, path_imgrec, 'r')
+
+i = 0
+while True:
+    try:
+        print(i)
+        header, s = recordio.unpack(imgrec.read_idx(i+1))
+        #print(str(header.label))
+        #img = np.array(mx.image.imdecode(s))
+        img = mx.image.imdecode(s).asnumpy()
+        #print(type(img))
+        path = os.path.join('images',str(header.label))
+        if not os.path.exists(path):
+                os.makedirs(path)
+        path = os.path.join(path,str(i))
+        #fig = plt.figure(frameon=False)
+        #fig.set_size_inches(124,124)
+        #ax = plt.Axes(fig, [0., 0., 1., 1.])
+        #ax.set_axis_off()
+        #fig.add_axes(ax)
+        #ax.imshow(img, aspect='auto')
+        #dpi=1
+        #fname= str(i)+'jpg'
+        #fig.savefig(fname, dpi)
+        #plt.savefig(path+'.jpg',bbox_inches='tight',pad_inches=0)
+        (b,g,r)=cv2.split(img)
+        img = cv2.merge([r,g,b])
+        #w,h = img.size
+        #print((img.shape))
+        cv2.imwrite(path+'.jpg',img)
+        i += 1
+    except EOFError:
+        break
+
+# 1~ 2369931.jpg 17gb
+```