Merge pull request #5 from QVPR/v1.1.0-alpha

V1.0.0 alpha to v1.0.0 full main merge

.gitignore

@@ -1,3 +1,4 @@
 weights/
+output/
 __pycache__/
 src/__pycache__/

src/blitnet.py

@@ -119,8 +119,6 @@ def addLayer(net,dims,thr_range,fire_rate,ip_rate,const_inp,nois,rec_spks):
     net['spikes'].append(torch.empty([],dtype=torch.float64))
     net['rec_spks'].append(rec_spks)
 
-    return len(net['x'])-1
-
 ##################################
 # Add a set of random connections between layers
 #  net:        BITnet instance
@@ -196,8 +194,6 @@ def addWeights(net,layer_pre,layer_post,W_range,p,stdp_rate):
         nrmInh[nrmInh==0.0] = 1.0
         net['W'][excIndex][n] = net['W'][excIndex][n,:,:]/nrmExc
         net['W'][inhIndex][n] = net['W'][inhIndex][n,:,:]/nrmInh
-
-    return len(net['W'])-1
 
 ##################################
 # Normalise all the firing rates
@@ -312,9 +308,7 @@ def calc_spikes(net,layersInfo):
     for i,eta in enumerate(net['eta_ip']):
 
         if net['rec_spks'][i]:
-            outspk = (net['x'][i][0,:,:]).detach().cpu().numpy() # detach to numpy for easy plotting
-            if i == 2:
-                outspk[outspk<0.05] = 0
+            outspk = (net['x'][i][0,0,:]).detach().cpu().numpy() # detach to numpy for easy plotting
             n_idx = np.nonzero(outspk)
             net['spikes'][i].extend([net['step_num']+net['x'][i][0,:,:].detach().cpu().numpy(),n]
                                         for n in n_idx)

src/metrics.py

@@ -53,14 +53,14 @@ def createPR(S_in, GThard, GTsoft=None, matching='multi', n_thresh=100):
 
     # single-best-match or multi-match VPR
     if matching == 'single':
-        # count the number of ground-truth positives (GTP)
-        GTP = np.count_nonzero(GT.any(0))
-
         # GT-values for best match per query (i.e., per column)
-        GT = GT[np.argmax(S, axis=0), np.arange(GT.shape[1])]
+        GT = GT[np.nanargmax(S, axis=1), np.arange(GT.shape[1])]
+
+         # count the number of ground-truth positives (GTP)
+        GTP = np.count_nonzero(GT)
 
         # similarities for best match per query (i.e., per column)
-        S = np.max(S, axis=0)
+        S = np.nanmax(S, axis=0)
 
     elif matching == 'multi':
         # count the number of ground-truth positives (GTP)

src/nordland.py

@@ -0,0 +1,112 @@
+'''
+Imports
+'''
+import os
+import re
+import shutil
+import zipfile
+
+from os import walk
+
+# load and sort the file names in order, not by how OS indexes them
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+def natural_keys(text):
+    return [ atoi(c) for c in re.split(r'(\d+)', text) ]
+
+# set the base path to the location of the downloaded Nordland datasets
+basePath = ''
+assert(os.path.isdir(basePath)),"Please set the basePath to the location of the downloaded Nordland datasets"
+
+# define the subfolders of the Nordland datasets
+subPath = ["spring_images_train/section1/","spring_images_train/section2/",
+           "fall_images_train/section1/","fall_images_train/section2/",
+           "winter_images_train/section1/","winter_images_train/section2/",
+           "summer_images_train/section1/","summer_images_train/section2/"]
+
+# set the desired output folder for unzipping and organization
+outDir = ''
+assert(os.path.isdir(outDir)),"Please set the outDir to the desired output location for unzipping the Nordland datasets"
+
+# define output paths for the data
+outPath = [os.path.join(outDir,"spring/"),os.path.join(outDir,"fall/"),
+           os.path.join(outDir,"winter/"),os.path.join(outDir,"summer/")]
+
+# check for existence of the zip folders, throw exception if missing
+zipNames = ["spring_images_train.zip","fall_images_train.zip",
+            "winter_images_train.zip","summer_images_train.zip"]
+for n in zipNames:
+    if not os.path.exists(basePath+n):
+        raise Exception('Please ensure dataset .zip folders have been downloaded') 
+
+# check if nordland data folders have already been unzipped
+zip_flag = []
+for n, ndx in enumerate(range(0,len(subPath),2)):
+    print('Unzipping '+zipNames[n])
+    if os.path.exists(basePath+subPath[ndx]):
+        # check if the folder contains any files
+        file_lst = os.listdir(basePath+subPath[ndx])
+        # remove folder if it is empty and unzip the data folder
+        if len(file_lst) == 0:
+            shutil.rmtree(basePath+subPath[ndx].replace('section1/',''))
+            with zipfile.ZipFile(basePath+zipNames[n],"r") as zip_ref:
+                zip_ref.extractall(basePath)
+    else:
+        with zipfile.ZipFile(basePath+zipNames[n],"r") as zip_ref:
+            zip_ref.extractall(basePath)
+
+# load image paths
+tempPaths = []
+imgPaths = []
+for n in range(0,len(subPath)):
+    tempPaths = []
+    for (path, dir_names, file_names) in walk(basePath+subPath[n]):
+        tempPaths.extend(file_names)   
+    # sort image names
+    tempPaths.sort(key=natural_keys)
+    tempPaths = [basePath+subPath[n]+s for s in tempPaths]
+    imgPaths = imgPaths + tempPaths
+
+# if output folders already exist, delete them
+for n in outPath:
+    if os.path.exists(n):
+        shutil.rmtree(n)
+        print('Removed pre-existing output folder')
+
+# rename and move the training data to match the nordland_imageNames.txt file
+for n in outPath:
+    os.mkdir(n)
+for n, filename in enumerate(imgPaths):
+    base = os.path.basename(filename)
+    split_base = os.path.splitext(base)
+    if int(split_base[0]) < 10:
+        my_dest = "images-0000"+split_base[0] + ".png"
+    elif int(split_base[0]) < 100:
+        my_dest = "images-000"+split_base[0] + ".png"
+    elif int(split_base[0]) < 1000:
+        my_dest = "images-00"+split_base[0] + ".png"
+    elif int(split_base[0]) < 10000:
+        my_dest = "images-0"+split_base[0] + ".png"
+    else:
+        my_dest = "images-"+split_base[0] + ".png"
+    if "spring" in filename:
+        out = outPath[0]
+    elif "fall" in filename:
+        out = outPath[1]
+    elif "winter" in filename:
+        out = outPath[2]
+    else:
+        out = outPath[-1]
+
+    fileDest = out + my_dest   
+    os.rename(filename, fileDest)
+
+# remove the empty folders
+for n, ndx in enumerate(subPath):
+    if n%2 == 0:
+        shutil.rmtree(basePath+ndx.replace('section1/',''))
+    else:
+        continue
+
+print('Finished unzipping and organizing Nordland dataset')