def main(args):
classifier = CamClassifier()
classifier.set_batch_size(1) # TODO: Batch if its running to slowly.
with open(args.dataset) as dataset:
labeled_patches = [line.split() for line in dataset]
labeled_patches = [(x[0], int(x[1])) for x in labeled_patches]
num_labels = {label: 0 for label in K_LABELS}
num_classified = {label: 0 for label in K_LABELS}
num_correct = {label: 0 for label in K_LABELS}
num_present = {label: 0 for label in K_LABELS}
confusion = {}
for label in K_LABELS:
confusion[label] = {}
for label2 in K_LABELS:
confusion[label][label2] = 0
counter = 0
for patch, label in labeled_patches:
print patch
counter += 1
# print counter
patch = cv2.imread(patch, cv2.IMREAD_COLOR)
label = K_LABELS[label]
if patch is None:
continue
classification = classify(classifier, patch)
confusion[classification][label] += 1
print label, classification
num_labels[label] += 1
num_classified[classification] += 1
if label == classification:
num_correct[classification] += 1
if classification != 'noise':
num_present[label] += 1
print num_correct
print num_classified
print num_labels
for label in K_LABELS:
num_classified[label] = max(1, num_classified[label])
num_labels[label] = max(1, num_labels[label])
precision = {l: num_correct[l] / float(num_classified[l]) for l in K_LABELS}
recall = {l: num_correct[l] / float(num_labels[l]) for l in K_LABELS}
for label in K_LABELS:
print ("%s\n\tprecision: %f\n\trecall: %f" % (label, precision[label], recall[label]))
num_present['animal'] = 0
total_present = sum(num_present.values())
for label in K_LABELS:
print "%s filter precentage: %f" % (label, num_present[label]/float(total_present))
print confusion
from detector.Webcam import Webcam
from classify.classifier import ReferenceClassifier
from classify.classifier import CamClassifier
parser = argparse.ArgumentParser()
parser.add_argument("frames_dir",
help = "Directory containing webcam image folders")
parser.add_argument("num_webcams",
help = "Number of webcams to sample", type = int)
args = parser.parse_args()
frames_dir = args.frames_dir
num_webcams = args.num_webcams
wc_paths = [os.path.join(frames_dir, fn) for fn in os.listdir(frames_dir)]
wc_paths = random.sample(wc_paths, num_webcams)
classifier = CamClassifier()
classifier.set_batch_size(1)
while True:
for path in wc_paths:
cam = Webcam(online=False, path=path)
for i in range(5):
cam.update()
overlay = cam.filtered_overlay(classifier, squareSize=227)
if overlay is not None:
cv2.imshow('overlay', overlay)
cv2.waitKey(0)
