(ง •̀_•́)ง we're heading for modeling!

Chunk
1 parent 5c9c44da
Showing 2 changed files with 71 additions and 1 deletions Show diff stats
mdata/ANALYSIS.py
mdata/crop.py
@@ -93,11 +93,28 @@ def anal_0000():
     df_ILS.hist(column='size',bins=100)
     plt.show()
+
+
+def pre_crop():
+    df_ILS = pd.read_csv('../res/file-tag-test.tsv', names=['hash', 'width', 'height', 'size', 'quality','chosen','class'], sep='\t')
+    print df_ILS.shape
+    print df_ILS[(df_ILS['width'] >= 300) & (df_ILS['height'] >= 300)].shape
+
+    # 300x300 4213 0.917 *
+    # 200x200 4534 0.987
+    # 400x400 932 0.202
+
+
+
+
 if __name__ == '__main__':
     # anal_ILSVRC()
     # anal_ILSVRC_Test()
-    anal_0000()
+    # anal_0000()
     # print timeit.timeit("anal_ILSVRC()", setup="from __main__ import anal_ILSVRC", number=1)
+
+
+    pre_crop()
     pass
@@ -0,0 +1,53 @@
+__author__ = 'chunk'
+
+import os, sys
+from PIL import Image
+from common import *
+import random
+
+base_dir = '/data/hadoop/ImageNet/ILSVRC/ILSVRC2013_DET_val/'
+category = 'Test'
+
+
+def crop_Test():
+    for path, subdirs, files in os.walk(os.path.join(base_dir, category)):
+        for name in files:
+            image = os.path.join(path, name)
+            im = Image.open(image)
+            w, h = im.size
+            if w < 300 or h < 300:
+                continue
+            left, upper = random.randint(0, w - 300), random.randint(0, h - 300)
+            im = im.crop((left, upper, left + 300, upper + 300))
+            im.save(os.path.join(base_dir, category + '_crop', name))
+
+
+
+if __name__ == '__main__':
+    timer = Timer()
+
+    timer.mark()
+    crop_Test()
+    timer.report()
+
+    pass
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
	@@ -93,11 +93,28 @@ def anal_0000():		@@ -93,11 +93,28 @@ def anal_0000():
93	df_ILS.hist(column='size',bins=100)	93	df_ILS.hist(column='size',bins=100)
94	plt.show()	94	plt.show()
95		95
		96	+
		97	+
		98	+def pre_crop():
		99	+ df_ILS = pd.read_csv('../res/file-tag-test.tsv', names=['hash', 'width', 'height', 'size', 'quality','chosen','class'], sep='\t')
		100	+ print df_ILS.shape
		101	+ print df_ILS[(df_ILS['width'] >= 300) & (df_ILS['height'] >= 300)].shape
		102	+
		103	+ # 300x300 4213 0.917 *
		104	+ # 200x200 4534 0.987
		105	+ # 400x400 932 0.202
		106	+
		107	+
		108	+
		109	+
96	if __name__ == '__main__':	110	if __name__ == '__main__':
97	# anal_ILSVRC()	111	# anal_ILSVRC()
98	# anal_ILSVRC_Test()	112	# anal_ILSVRC_Test()
99	- anal_0000()	113	+ # anal_0000()
100	# print timeit.timeit("anal_ILSVRC()", setup="from __main__ import anal_ILSVRC", number=1)	114	# print timeit.timeit("anal_ILSVRC()", setup="from __main__ import anal_ILSVRC", number=1)
		115	+
		116	+
		117	+ pre_crop()
101	pass	118	pass
102		119
103		120
@@ -0,0 +1,53 @@		@@ -0,0 +1,53 @@
	1	+__author__ = 'chunk'
	2	+
	3	+import os, sys
	4	+from PIL import Image
	5	+from common import *
	6	+import random
	7	+
	8	+base_dir = '/data/hadoop/ImageNet/ILSVRC/ILSVRC2013_DET_val/'
	9	+category = 'Test'
	10	+
	11	+
	12	+def crop_Test():
	13	+ for path, subdirs, files in os.walk(os.path.join(base_dir, category)):
	14	+ for name in files:
	15	+ image = os.path.join(path, name)
	16	+ im = Image.open(image)
	17	+ w, h = im.size
	18	+ if w < 300 or h < 300:
	19	+ continue
	20	+ left, upper = random.randint(0, w - 300), random.randint(0, h - 300)
	21	+ im = im.crop((left, upper, left + 300, upper + 300))
	22	+ im.save(os.path.join(base_dir, category + '_crop', name))
	23	+
	24	+
	25	+
	26	+if __name__ == '__main__':
	27	+ timer = Timer()
	28	+
	29	+ timer.mark()
	30	+ crop_Test()
	31	+ timer.report()
	32	+
	33	+ pass
	34	+
	35	+
	36	+
	37	+
	38	+
	39	+
	40	+
	41	+
	42	+
	43	+
	44	+
	45	+
	46	+
	47	+
	48	+
	49	+
	50	+
	51	+
	52	+
	53	+