sparrow.py

   1 import struct
   2 import numpy as np
   3 import heapq
   4 from itertools import count
   5
   6
   7 INDEX_FILE = 'content/jpeg.index'
   8 TEXT_INDEX_FILE = 'content/jpeg.index.txt'
   9 FRAME_STRUCTURE = 'II48s8I'
  10 BLOB_NAME='content/jpeg.blob'
  11
  12
  13 def save_frames(frames, filename):
  14     f = open(filename, 'w')
  15     for frame in frames:
  16         f.write(struct.pack(FRAME_STRUCTURE,
  17                             frame.glob_index,
  18                             frame.jpeg_len,
  19                             frame.summary,
  20                             *frame.successors
  21                             ))
  22
  23
  24 def save_frames_text(frames, filename):
  25     f = open(filename, 'w')
  26     for frame in frames:
  27         print >> f, ("Frame:")
  28         print >> f, (" index: %d" % frame.index)
  29         print >> f, (" glob_index: %d" % frame.glob_index)
  30         print >> f, (" jpeg_len: %d" % frame.jpeg_len)
  31         print >> f, (" summary: %s" % list(frame.summary))
  32         print >> f, (" successors: %s" % frame.successors)
  33
  34
  35 class Frame:
  36     def __init__(self, packed=None):
  37         if packed is not None:
  38             data = struct.unpack(FRAME_STRUCTURE, packed)
  39             self.glob_index = data[0]
  40             self.jpeg_len = data[1]
  41             self.summary = data[2]
  42             self.array = np.fromstring(data[2], dtype=np.uint8)
  43             self.successors = list(data[3:])
  44             self.next = self.successors[0]
  45
  46
  47
  48 def distance_gen(tail, heads):
  49     s = tail.array
  50     for h in heads:
  51         if h is not tail.head:
  52             yield((sum((s - h.array) ** 2), h))
  53
  54
  55 def link_frames(frames):
  56     tails = []
  57     heads = []
  58     isbreak = True
  59     for f in frames:
  60         if isbreak:
  61             heads.append(f)
  62             isbreak = False
  63         if f.successors[0] == 0:
  64             f.head = heads[-1]
  65             tails.append(f)
  66             isbreak = True
  67
  68     for f in heads + tails:
  69         if not hasattr(f, 'array'):
  70             f.array =  np.fromstring(f.summary, dtype=np.uint8)
  71
  72     #print frames, heads, tails
  73
  74     for t in tails:
  75         closest = [x[1] for x in heapq.nsmallest(7, distance_gen(t, heads), key=lambda x: x[0])]
  76         t.successors = [0] + [x.index for x in closest]
  77         print t.index, t.successors
  78
  79     return frames