Real-time yolo

Author: saimj7

Object detection using YOLO (DarkNet)/cam.py

@@ -0,0 +1,51 @@
+import cv2
+from darkflow.net.build import TFNet
+import numpy as np
+import time #how fast is the processing
+
+options = {
+    'model': 'cfg/yolo.cfg', #model in cfg folder with yolo model
+    'load': 'bin/yolo.weights', #loading weights from bin folder downloaded from darknet
+    'threshold': 0.2, #confidence factor to draw bounding box, > 0.2 to show in image, low = tons of boxes which is not good
+    'gpu': 1.0 #uses GPU to render images
+}
+
+tfnet = TFNet(options) #object called tfnet to run options
+#displays info in command window like trained, layer description, output size
+
+colors = [tuple(255 * np.random.rand(3)) for _ in range(10)] # list of random colors for variations to boxes, array with 3 elements long
+# for in range(10) = sicne we want few entries; we wanna go from 0 t0 255
+
+#0 for webcam 1, diff no for multiple cams
+capture = cv2.VideoCapture(0)
+capture.set(cv2.CAP_PROP_FRAME_WIDTH, 1920) #full hd
+capture.set(cv2.CAP_PROP_FRAME_HEIGHT, 1080)
+
+while True:
+    stime = time.time() #start time
+    ret, frame = capture.read() # ret is true while playing, false when ends; frame is actual frame; reading from capture device
+    if ret:
+        results = tfnet.return_predict(frame) # return_predict operation predicts whats in the video frame; passing the frame
+
+        # define bounding box and display label on the frame
+        #looping over all items, all results, adding box for each thing it detects; loops over all predictions
+        for color, result in zip(colors, results): # zip makes a list and then each item is gonna be a tuple with one color and one result
+            #create tuples with two corners x,y in it
+            tl = (result['topleft']['x'], result['topleft']['y']) #take first element as top left from result
+            br = (result['bottomright']['x'], result['bottomright']['y']) #bottom right
+            label = result['label']
+            confidence = result['confidence']
+            text = '{}: {:.0f}%'.format(label, confidence * 100) # string, float with 0 decimals; format with label, conf and mul with 100 to get %
+
+            #redefine frame with cv2 rectangle, pass frame first, then tl,br,color we want, line width
+            frame = cv2.rectangle(frame, tl, br, color, 5)
+            #add label by redifing frame with cv2 puttext; pass img first, text=label, font, font size, color, line width
+            frame = cv2.putText( #puttext on the frame
+                frame, text, tl, cv2.FONT_HERSHEY_COMPLEX, 1, (0, 0, 0), 2)
+        cv2.imshow('frame', frame) #display frame
+        print('FPS {:.1f}'.format(1 / (time.time() - stime))) #.1f = one decimal place, calculation
+    if cv2.waitKey(1) & 0xFF == ord('q'): #boilder plate code, if hit q key it ends
+        break
+
+capture.release()
+cv2.destroyAllWindows()

Object detection using YOLO (DarkNet)/yolo.cfg

@@ -0,0 +1,258 @@
+[net]
+# Testing
+batch=1
+subdivisions=1
+# Training
+# batch=64
+# subdivisions=8
+width=608
+height=608
+channels=3
+momentum=0.9
+decay=0.0005
+angle=0
+saturation = 1.5
+exposure = 1.5
+hue=.1
+
+learning_rate=0.001
+burn_in=1000
+max_batches = 500200
+policy=steps
+steps=400000,450000
+scales=.1,.1
+
+[convolutional]
+batch_normalize=1
+filters=32
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[maxpool]
+size=2
+stride=2
+
+[convolutional]
+batch_normalize=1
+filters=64
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[maxpool]
+size=2
+stride=2
+
+[convolutional]
+batch_normalize=1
+filters=128
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=64
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=128
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[maxpool]
+size=2
+stride=2
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=128
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[maxpool]
+size=2
+stride=2
+
+[convolutional]
+batch_normalize=1
+filters=512
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=512
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=256
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=512
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[maxpool]
+size=2
+stride=2
+
+[convolutional]
+batch_normalize=1
+filters=1024
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=512
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=1024
+size=3
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=512
+size=1
+stride=1
+pad=1
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+filters=1024
+size=3
+stride=1
+pad=1
+activation=leaky
+
+
+#######
+
+[convolutional]
+batch_normalize=1
+size=3
+stride=1
+pad=1
+filters=1024
+activation=leaky
+
+[convolutional]
+batch_normalize=1
+size=3
+stride=1
+pad=1
+filters=1024
+activation=leaky
+
+[route]
+layers=-9
+
+[convolutional]
+batch_normalize=1
+size=1
+stride=1
+pad=1
+filters=64
+activation=leaky
+
+[reorg]
+stride=2
+
+[route]
+layers=-1,-4
+
+[convolutional]
+batch_normalize=1
+size=3
+stride=1
+pad=1
+filters=1024
+activation=leaky
+
+[convolutional]
+size=1
+stride=1
+pad=1
+filters=425
+activation=linear
+
+
+[region]
+anchors =  0.57273, 0.677385, 1.87446, 2.06253, 3.33843, 5.47434, 7.88282, 3.52778, 9.77052, 9.16828
+bias_match=1
+classes=80
+coords=4
+num=5
+softmax=1
+jitter=.3
+rescore=1
+
+object_scale=5
+noobject_scale=1
+class_scale=1
+coord_scale=1
+
+absolute=1
+thresh = .1
+random=1