config changes

Edmond Yoo

2018-09-14 0c41868cc815cd4b15fbbbd8aed9156e49a0627a

config changes

 .gitignore

@@ -5,7 +5,6 @@
*.png
*.so
mnist/
data/
caffe/
grasp/
images/

 cfg/tiny_yolo.cfg

@@ -119,7 +119,7 @@
bias_match=1
classes=1
coords=4
num=5
num=9
softmax=1
jitter=.2
rescore=1

 cfg/yolov3-custom.cfg

@@ -608,7 +608,7 @@
mask = 6,7,8
anchors = 129.1513,150.1077, 104.9181,195.1675, 145.8067,186.4670, 127.9706,231.2296, 165.9494,221.8162, 162.4676,253.0225
classes=1
num=9
num=6
jitter=.3
ignore_thresh = .7
truth_thresh = 1
@@ -694,7 +694,7 @@
mask = 3,4,5
anchors = 129.1513,150.1077, 104.9181,195.1675, 145.8067,186.4670, 127.9706,231.2296, 165.9494,221.8162, 162.4676,253.0225
classes=1
num=9
num=6
jitter=.3
ignore_thresh = .7
truth_thresh = 1
@@ -781,7 +781,7 @@
mask = 0,1,2
anchors = 129.1513,150.1077, 104.9181,195.1675, 145.8067,186.4670, 127.9706,231.2296, 165.9494,221.8162, 162.4676,253.0225
classes=1
num=9
num=6
jitter=.3
ignore_thresh = .7
truth_thresh = 1

 darknet.py

@@ -32,6 +32,7 @@
import random
import os
import cv2
import numpy as np

def sample(probs):
    s = sum(probs)
@@ -223,7 +224,10 @@
    Performs the meat of the detection
    """
    #pylint: disable= C0321
    im = load_image(image, 0, 0)
    if isinstance(image, np.ndarray):
        im = array_to_image(image)[0]
    else:
        im = load_image(image, 0, 0)
    #import cv2
    #custom_image_bgr = cv2.imread(image) # use: detect(,,imagePath,)
    #custom_image = cv2.cvtColor(custom_image_bgr, cv2.COLOR_BGR2RGB)
@@ -367,7 +371,8 @@
        raise ValueError("Invalid image path `"+os.path.abspath(imagePath)+"`")
    # Do the detection
    #detections = detect(netMain, metaMain, imagePath, thresh)  # if is used cv2.imread(image)
    detections = detect(netMain, metaMain, imagePath.encode("ascii"), thresh)
    #detections = detect(netMain, metaMain, imagePath.encode("ascii"), thresh)
    detections = detect(netMain, metaMain, cv2.imread(imagePath), thresh, debug=True)
    if showImage:
        try:
            from skimage import io, draw
@@ -423,7 +428,7 @@
    return detections


def capture(thresh=.5, hier_thresh=.5, nms=.45, configPath="./cfg/yolov3.cfg", weightPath="yolov3.weights",
def capture(thresh=.25, hier_thresh=.5, nms=.45, configPath="./cfg/yolov3.cfg", weightPath="yolov3.weights",
            metaPath="./data/coco.data", showImage=True, makeImageOnly=False, initOnly=False):
    global metaMain, netMain, altNames  # pylint: disable=W0603
    netMain = load_net_custom(configPath.encode("ascii"), weightPath.encode("ascii"), 0, 1)  # batch size = 1
@@ -433,7 +438,7 @@
    pnum = pointer(num)
    num = pnum[0]

    capture = cv2.VideoCapture('../data/test3.mp4')
    capture = cv2.VideoCapture('../data/test1.mp4')
    print(capture.get(cv2.CAP_PROP_FPS))

    capture.set(cv2.CAP_PROP_FRAME_WIDTH, 1024)
@@ -441,6 +446,8 @@

    while True:
        ret, frame = capture.read()
        detections = detect(netMain, metaMain, frame, thresh, debug=True)
        '''
        im, arr = array_to_image(frame)
        predict_image(netMain, im)
        dets = get_network_boxes(netMain, im.w, im.h, thresh, hier_thresh, None, 0, pnum, 1)
@@ -453,6 +460,42 @@
                    b = dets[j].bbox
                    nameTag = metaMain.names[i]
                    res.append((nameTag, dets[j].prob[i], (b.x, b.y, b.w, b.h)))
        '''
        for detection in detections:
            label = detection[0]
            confidence = detection[1]
            pstring = label + ": " + str(np.rint(100 * confidence)) + "%"
            imcaption.append(pstring)
            print(pstring)
            bounds = detection[2]
            shape = image.shape
            # x = shape[1]
            # xExtent = int(x * bounds[2] / 100)
            # y = shape[0]
            # yExtent = int(y * bounds[3] / 100)
            yExtent = int(bounds[3])
            xEntent = int(bounds[2])
            # Coordinates are around the center
            xCoord = int(bounds[0] - bounds[2] / 2)
            yCoord = int(bounds[1] - bounds[3] / 2)
            boundingBox = [
                [xCoord, yCoord],
                [xCoord, yCoord + yExtent],
                [xCoord + xEntent, yCoord + yExtent],
                [xCoord + xEntent, yCoord]
            ]
            # Wiggle it around to make a 3px border
            rr, cc = draw.polygon_perimeter([x[1] for x in boundingBox], [x[0] for x in boundingBox], shape=shape)
            rr2, cc2 = draw.polygon_perimeter([x[1] + 1 for x in boundingBox], [x[0] for x in boundingBox], shape=shape)
            rr3, cc3 = draw.polygon_perimeter([x[1] - 1 for x in boundingBox], [x[0] for x in boundingBox], shape=shape)
            rr4, cc4 = draw.polygon_perimeter([x[1] for x in boundingBox], [x[0] + 1 for x in boundingBox], shape=shape)
            rr5, cc5 = draw.polygon_perimeter([x[1] for x in boundingBox], [x[0] - 1 for x in boundingBox], shape=shape)
            boxColor = (int(255 * (1 - (confidence ** 2))), int(255 * (confidence ** 2)), 0)
            draw.set_color(image, (rr, cc), boxColor, alpha=0.8)
            draw.set_color(image, (rr2, cc2), boxColor, alpha=0.8)
            draw.set_color(image, (rr3, cc3), boxColor, alpha=0.8)
            draw.set_color(image, (rr4, cc4), boxColor, alpha=0.8)
            draw.set_color(image, (rr5, cc5), boxColor, alpha=0.8)
        print(res)
        cv2.imshow('frame', frame)
        if cv2.waitKey(1) & 0xFF == ord('q'):
@@ -463,8 +506,9 @@


if __name__ == "__main__":
    performDetect(imagePath="../data/test1.jpg", thresh=0.25, configPath="./cfg/tiny_yolo.cfg",
                  weightPath="./weights/second_general/tiny_yolo_17000.weights",
                  metaPath="./data/obj.data", showImage=True, makeImageOnly=False, initOnly=False)
    performDetect(imagePath='data/scream.jpg')
    #performDetect(imagePath="../data/test1.jpg", thresh=0.25, configPath="./cfg/tiny_yolo.cfg",
    #              weightPath="./weights/second_general/tiny_yolo_17000.weights",
    #              metaPath="./data/obj.data", showImage=True, makeImageOnly=False, initOnly=False)
    #print(performDetect(showImage=False))
    #capture()

 data/coco.data

 data/dog.txt

New file
@@ -0,0 +1,3 @@
16 0.2929 0.6516 0.2627 0.5270
7 0.7453 0.2198 0.3476 0.1569
1 0.4540 0.4864 0.5653 0.5323

 data/obj.data

 data/obj.names