adding the data

busProjectTest.py

@@ -0,0 +1,190 @@
+import os
+import sys
+import matplotlib
+import matplotlib.pyplot as plt
+import matplotlib.patches as patches
+import numpy as np
+import ast
+
+import imageio
+
+
+fontdict = {'fontsize':15, 'weight':'bold'}
+plt.switch_backend('Qt5Agg')
+
+class IMAGE:
+
+    def __init__(self):
+        self.ROIS = []
+
+    def set_image(self, im):
+        self.I = imageio.imread(im)
+        self.imShape = self.I.shape
+
+    def clear_ROIS(self):
+        self.ROIS = []
+
+    def add_ROI(self, pos):
+        self.ROIS.append(pos)
+
+    def show_ROI(self, title, edgecolor, numGT, text, saveDir = None):
+        fig, ax = plt.subplots(1)
+        ax.imshow(self.I)
+        if(not isinstance(edgecolor,list) and len(self.ROIS) > 0):
+            edgecolor = [edgecolor] * len(self.ROIS)
+        for i in range(0,numGT):
+            ROI = self.ROIS[i]
+            rect = patches.Rectangle((ROI[0], ROI[1]), ROI[2], ROI[3], linewidth = 1, edgecolor = edgecolor[i], facecolor='none')
+            ax.add_patch(rect)
+        for i in range(numGT,len(self.ROIS)):
+            ROI = self.ROIS[i]
+            rect = patches.Rectangle((ROI[0], ROI[1]), ROI[2], ROI[3], linewidth = 1, edgecolor = edgecolor[i], facecolor='none', linestyle = '--')
+            ax.add_patch(rect)
+        if(saveDir is None):
+            ax.text(15,160,text, fontdict = fontdict, bbox={'facecolor':'yellow', 'edgecolor':'yellow','alpha':0.5, 'pad':2})
+        else:
+            ax.text(15, 300, text, fontdict=fontdict,bbox={'facecolor': 'yellow', 'edgecolor': 'yellow', 'alpha': 0.5, 'pad': 2})
+        plt.title(title)
+        if(not saveDir is None):
+            plt.savefig(os.path.join(saveDir, title), dpi = 500)
+        plt.close()
+
+    def close(self):
+        plt.close()
+
+def IOU(boxAList, boxBList):
+    Th = 0.69
+    iou = []
+    matches = {}
+    tp = 0
+    fp = len(boxBList)
+    missed = len(boxAList)
+    for i in range(len(boxAList)):
+        boxA = boxAList[i][:4]
+        iou_ = []
+        for j in range(len(boxBList)):
+            boxB = boxBList[j][:4]
+            if(not ((boxB[0] <= boxA[0] <= boxB[0] + boxB[2]) or (boxA[0] <= boxB[0] <= boxA[0] + boxA[2]))):
+                iou_.append(0.0)
+                continue
+            xA = max(boxA[0], boxB[0])
+            yA = max(boxA[1], boxB[1])
+            xB = min(boxA[0] + boxA[2], boxB[0] + boxB[2])
+            yB = min(boxA[1] + boxA[3], boxB[1] + boxB[3])
+            interArea = (xB - xA + 1) * (yB - yA + 1)
+            boxAArea = (boxA[2] + 1)*(boxA[3] + 1)
+            boxBArea = (boxB[2] + 1)*(boxB[3] + 1)
+            iou_.append(interArea / float(boxAArea + boxBArea - interArea))
+        maxIou = max(iou_)
+        maxIouIndex = iou_.index(max(iou_))
+        iou.append(maxIou)
+        if (maxIouIndex in matches and maxIou > iou[matches[maxIouIndex]]):
+            if (iou[matches[maxIouIndex]] > Th and boxAList[matches[maxIouIndex]][4] == boxBList[maxIouIndex][4]):
+                pass
+            elif(maxIou > Th and boxAList[i][4] == boxBList[maxIouIndex][4]):
+                tp += 1
+                missed -= 1
+                fp -= 1
+            matches[maxIouIndex] = i
+        if(not maxIouIndex in matches):
+            matches[maxIouIndex] = i
+            if(maxIou > Th and boxAList[i][4] == boxBList[maxIouIndex][4]):
+                tp += 1
+                missed -= 1
+                fp -= 1
+    return tp, fp, missed, iou
+
+def runTest(annFileNameGT, myAnnFileName, busDir , saveDir = None, elapsed = None):
+
+    image = IMAGE()
+    objectsColors = {'g':'1', 'y':'2', 'w':'3', 's':'4', 'b':'5', 'r':'6'}
+    objectsColorsInv = {v: k for k, v in objectsColors.items()}
+    objectsColorsForShow = {'g':'g', 'y':'y', 'w':'w', 's':'tab:gray', 'b':'b', 'r':'r'}
+
+    writtenAnnsLines = {}
+    annFileEstimations = open(myAnnFileName, 'r')
+    annFileGT = open(annFileNameGT, 'r')
+    writtenAnnsLines['Ground_Truth'] = (annFileGT.readlines())
+    writtenAnnsLines['Estimation'] = (annFileEstimations.readlines())
+
+    TP = 0
+    FP = 0
+    MISS = 0
+
+    for i in range(len(writtenAnnsLines['Ground_Truth'])):
+
+        lineGT = writtenAnnsLines['Ground_Truth'][i].replace(' ','')
+        colors = []
+        imName = lineGT.split(':')[0]
+        lineE = [x for x in writtenAnnsLines['Estimation'] if imName == x.split(':')[0]]
+        if(len(lineE) == 0):
+            lineE = imName + ':'
+        else:
+            lineE = lineE[0]
+        bus = os.path.join(busDir, imName)
+        image.set_image(bus)
+        image.clear_ROIS()
+        annsGT = lineGT[lineGT.index(':') + 1:].replace('\n', '')
+        annsE = lineE[lineE.index(':') + 1:].replace('\n', '')
+        annsGT = ast.literal_eval(annsGT)
+        if (not isinstance(annsGT, tuple)):
+            annsGT = [annsGT]
+        for ann in annsGT:
+            image.add_ROI(ann[:4])
+            colorTag = objectsColorsInv[str(ann[4])]
+            colors.append(objectsColorsForShow[colorTag])
+        numGT = len(annsGT)
+        if('[' in lineE):
+            annsE = ast.literal_eval(annsE)
+            if (not isinstance(annsE, tuple)):
+                annsE = [annsE]
+            for ann in annsE:
+                image.add_ROI(ann[:4])
+                colorTag = objectsColorsInv[str(ann[4])]
+                colors.append(objectsColorsForShow[colorTag])
+            tp, fp, missed, iou = IOU(annsGT, annsE)
+        else:
+            tp = 0
+            fp = 0
+            numGT = 0
+            missed = len(annsGT)
+            iou = []
+        TP += tp
+        FP += fp
+        MISS += missed
+        print('Image : {}, TP: {} FP: {} MISSED : {}'.format(imName, tp, fp, missed))
+        iouStr = ','.join(['{0:.2f}'.format(x) for x in iou])
+        text = 'IOU Scores : ' + iouStr + '\nTP = {}, FP = {}, Missed = {} '.format(tp, fp, missed)
+        image.show_ROI(edgecolor = colors, title = imName, numGT = numGT , text = text, saveDir = saveDir)
+
+    if(TP == 0):
+        F1Score = 0
+    else:
+        precision = TP/(TP + FP)
+        recall = TP/(TP + MISS)
+        F1Score = 2*(precision * recall)/(precision + recall)
+    strToWrite = 'Total detections = {}/{}\nTotal False Positives = {}\nTotal missed = {}'.format(TP, TP+MISS, FP, MISS)
+    strToWrite += '\nF1 SCORE : {0:.3f}'.format(F1Score)
+    if(not elapsed is None):
+        strToWrite += '\nTime elapsed : {0:.2f} seconds'.format(elapsed)
+    fig, ax = plt.subplots(1)
+    plt.title('Results', fontdict = {'fontsize':20})
+    im = np.zeros((10,26,3), dtype=np.uint8)
+    im[:,:,2] = 221
+    im[:,:,1] = 114
+    im[:,:,0] = 102
+    ax.imshow(im)
+    ax.text(4,7, strToWrite, style='italic', fontdict = {'fontsize':50, 'weight':'bold'})
+    mng = plt.get_current_fig_manager()
+    mng.window.showMaximized()
+    plt.show()
+    plt.close()
+    FIG, ax = plt.subplots(1)
+    plt.title('Results', fontdict = {'fontsize':20})
+    ax.imshow(im)
+    ax.text(2,8, strToWrite, style='italic', fontdict = {'fontsize':20, 'weight':'bold'})
+    if(saveDir is None):
+        saveDir = os.path.join(os.getcwd(), 'Output')
+        if(not os.path.exists(saveDir)):
+            os.mkdir(saveDir)
+    plt.savefig(os.path.join(saveDir,'Results.png'), dpi = 600)