Port to python 3

datasets.py

@@ -4,7 +4,8 @@
 import numpy as np
 import time
 import multiprocessing as mp
-import queue
+from multiprocessing import Queue
+from queue import Empty
 
 SIZE_INC = int(2048)
 CHUNK_SIZE = int(128)
@@ -39,12 +40,12 @@ def init_ds(self):
 
     def run(self):
         self.init_ds()
-        f = file('datasets_ioerrors.txt', 'a')
+        f = open('datasets_ioerrors.txt', 'a')
         while not self.exit.is_set() or not self.q.empty():
             try:
                 res = self.q.get(False, 1e-3)
                 self._save(res)
-            except Queue.Empty:
+            except Empty:
                 pass
             except IOError:
                 print('IOError, continuing')
@@ -57,7 +58,7 @@ def run(self):
         self.close()
 
     def create_datasets(self, tables, compression=None):
-        for tname, ttype in tables.iteritems():
+        for tname, ttype in tables.items():
             tname_split = tname.split('/')
             if len(tname_split) > 1:
                 grpname, subtname = tname_split
@@ -92,7 +93,7 @@ def save(self, data):
             raise Queue.Full('dataset buffer overflow')
 
     def _save(self, data):
-        for col,val in data.iteritems():
+        for col,val in data.items():
             self.outbuffers[col].append(val)
             if len(self.outbuffers[col]) == self.chunk_size:
                 self[col][self.ptrs[col]:self.ptrs[col] + self.chunk_size] = \

export.py

@@ -10,7 +10,8 @@
 
 from __future__ import print_function
 import os, sys, time, argparse
-import Queue
+from multiprocessing import Queue
+from queue import Empty
 import numpy as np
 import h5py
 from copy import deepcopy
@@ -133,7 +134,7 @@ def raster_evts(data):
     while m.has_data and sys_ts <= tstop*1e-6:
         try:
             sys_ts, d = m.get()
-        except Queue.Empty:
+        except Empty:
             # wait for queue to fill up
             time.sleep(0.01)
             continue
@@ -158,7 +159,7 @@ def raster_evts(data):
                 while t_pre + args.binsize < d['timestamp'] + t_offset:
                     # aps frame is not in current bin -> save and proceed
                     f_out.save(deepcopy(current_row))
-                    current_row['dvs_frame'][:,:] = 0
+                    current_row['dvs_frame'] = 0
                     current_row['timestamp'] = t_pre
                     t_pre += args.binsize
             else:
@@ -175,7 +176,7 @@ def raster_evts(data):
             if fixed_dt:
                 # fixed time interval bin mode
                 num_samples = int(np.ceil((times[-1] - t_pre) / args.binsize))
-                for _ in xrange(num_samples):
+                for _ in range(0, num_samples):
                     # take n events
                     n = (times[offset:] < t_pre + args.binsize).sum()
                     sel = slice(offset, offset + n)
@@ -191,7 +192,7 @@ def raster_evts(data):
             else:
                 # fixed event count mode
                 num_samples = np.ceil(-float(num_evts + ev_count)/args.binsize)
-                for _ in xrange(int(num_samples)):
+                for _ in range(int(num_samples)):
                     n = min(int(-args.binsize - ev_count), num_evts - offset)
                     sel = slice(offset, offset + n)
                     current_row['dvs_frame'] += raster_evts(d['data'][sel])

get_stats.py

@@ -30,7 +30,7 @@
         data = f[k]['data'][nnz, :]
         t, v = data[:, 0] * 1e-6, data[:, 1]
         curve = ip.interp1d(t, v)
-        tnew = np.linspace(t[0], t[-1], (t[-1] - t[0]) / tstep)
+        tnew = np.linspace(t[0], t[-1], int((t[-1] - t[0]) / tstep))
         vout.append(curve(tnew))
 
     out = np.hstack(vout)

hdf5_deeplearn_utils.py

@@ -1,6 +1,7 @@
 import h5py
 import numpy as np
-from scipy.misc import imresize
+#from scipy.misc import imresize
+from skimage.transform import resize
 from collections import defaultdict
 
 def get_real_endpoint(h5f):
@@ -15,7 +16,7 @@ def chunker(seq, size):
     return (seq[pos:pos + size] for pos in xrange(0, len(seq), size))
 
 def yield_chunker(seq, size):
-    for pos in xrange(0, len(seq), size):
+    for pos in range(0, len(seq), size):
         yield seq[pos:pos + size]
 
 def check_and_fix_timestamps(h5f):
@@ -237,11 +238,13 @@ def flow(self, h5fs, dataset_keys, indexes_key, batch_size, shuffle=True):
             b += 1
 
 def resize_int8(frame, size):
-    return imresize(frame, size)
+    #return imresize(frame, size)
+    return resize(frame, size)
 
 def resize_int16(frame, size=(60,80), method='bilinear', climit=[-15,15]):
     # Assumes data is some small amount around the mean, i.e., DVS event sums
-    return imresize((np.clip(frame, climit[0], climit[1]).astype('float32')+127), size, interp=method).astype('uint8')
+    #return imresize((np.clip(frame, climit[0], climit[1]).astype('float32')+127), size, interp=method).astype('uint8')
+    return resize((np.clip(frame, climit[0], climit[1]).astype('float32')+127), size).astype('uint8')
 
 def resize_data_into_new_key(h5f, key, new_key, new_size, chunk_size=1024):
     chunk_generator = yield_chunker(h5f[key], chunk_size)
@@ -255,7 +258,7 @@ def resize_data_into_new_key(h5f, key, new_key, new_size, chunk_size=1024):
     elif key == 'dvs_frame':
         do_resize = resize_int16
     else:
-        raise AssertionError, 'Unknown data type'
+        raise AssertionError('Unknown data type')
     # Initialize a resizable dataset to hold the output
     resized_shape = (chunk_size,) + new_size
     max_shape = (h5f[key].shape[0],) + resized_shape[1:]

interfaces/caer.py

@@ -13,7 +13,7 @@
 import numpy as np
 import socket, struct
 import multiprocessing as mp
-import Queue
+from multiprocessing import Queue
 
 HOST = "127.0.0.1"
 PORT = 7777

interfaces/oxc.py

@@ -13,7 +13,7 @@
 import multiprocessing as mp
 import numpy as np
 from openxc.tools import dump as oxc
-import Queue
+from multiprocessing import Queue
 
 class Monitor(mp.Process):
     def __init__(self, bufsize=256):

lasagne_utils.py

@@ -1,6 +1,6 @@
 import os
 import lasagne
-import cPickle as pickle
+import _pickle as pickle
 import numpy as np
 import theano.tensor as T
 # Create 5D tensor type
@@ -19,12 +19,12 @@ def save_model(filename, suffix, model, log=None, announce=True, log_only=False)
     if not log_only:
         # Acquire Data
         data = lasagne.layers.get_all_param_values(model)
-        with open(param_filename, 'w') as f:
+        with open(param_filename, 'wb') as f:
             pickle.dump(data, f)
     # Generate log filename and dump
     if log is not None:
         log_filename = '%s.log' % (filename)
-        with open(log_filename, 'w') as f:
+        with open(log_filename, 'wb') as f:
             pickle.dump(log, f)
 
 def load_model(filename, model):

view.py

@@ -75,7 +75,7 @@ def __init__(self, filename, tables, bufsize=64):
     def run(self):
         while self.blocks_rem and not self.exit.is_set():
             blocks_read = 0
-            for k in self.blocks_rem.keys():
+            for k in list(self.blocks_rem):
                 if self.q[k].full():
                     time.sleep(1e-6)
                     continue
@@ -591,7 +591,7 @@ def caer_event_from_row(row):
                         help="Examples:\n"
                              "-s 50%% - play file starting at 50%%\n"
                              "-s 66s - play file starting at 66s")
-    parser.add_argument('--rotate', '-r', type=bool, default=False,
+    parser.add_argument('--rotate', '-r', type=bool, default=True,
                         help="Rotate the scene 180 degrees if True, "
                              "Otherwise False")
     args = parser.parse_args()