d[ord(c)] = 1
pred = predict(net, d)
d[ord(c)] = 0
c = s[-1]
while 1:
d[ord(c)] = 1
pred = predict(net, d)
d[ord(c)] = 0
pred = [pred[i] for i in range(256)]
ind = sample(pred)
c = chr(ind)
prob += math.log(pred[ind])
if len(tac) and tac[-1] == '.':
break
tac = tac + c
return (tac, prob)
def predict_tactics(net, s, n):
tacs = []
for i in range(n):
reset_rnn(net)
tacs.append(predict_tactic(net, s))
tacs = sorted(tacs, key=lambda x: -x[1])
return tacs
net = load_net("cfg/coq.test.cfg", "/home/pjreddie/backup/coq.backup", 0)
t = predict_tactics(net, "+++++\n", 10)
print(t)
Created on Thu Aug 9 21:51:14 2018
@author: julien
"""
import cv2
import os
from python import darknet as dn
# load the Neural Network and the meta
path = "./char/"
cfg = os.path.join(path, "yolov3_plate.cfg").encode()
weights = os.path.join(path, "yolov3_plate_4000_ok.weights").encode()
data = os.path.join(path, "obj_plate.data").encode()
net = dn.load_net(cfg, weights, 0)
meta = dn.load_meta(data)
path_v = "/home/jehl/darknet/char/img_char"
files = os.listdir(path_v)
for f in files:
img_path = os.path.join(path_v, f)
result, im = dn.detect(net, meta, img_path.encode(), thresh=0.1)
if len(result) != 0:
print("Recognize !!!!!!!!!!!!!!!!!!")
for p in result:
x1 = int(p[2][0] - p[2][2] / 2)
y1 = int(p[2][1] - p[2][3] / 2)
x2 = int(p[2][0] + p[2][2] / 2)
import time
import random
import colorsys
import numpy as np
from PIL import Image, ImageDraw, ImageFont
import cv2
import python.darknet as dn
# prepare YOLO
net = dn.load_net(str.encode("cfg/tiny-yolo.cfg"),
str.encode("weights/tiny-yolo.weights"), 0)
meta = dn.load_meta(str.encode("cfg/coco.data"))
# box colors
box_colors = None
def generate_colors(num_classes):
global box_colors
if box_colors != None and len(box_colors) > num_classes:
return box_colors
hsv_tuples = [(x / num_classes, 1., 1.) for x in range(num_classes)]
box_colors = list(map(lambda x: colorsys.hsv_to_rgb(*x), hsv_tuples))
box_colors = list(
map(lambda x: (int(x[0] * 255), int(x[1] * 255), int(x[2] * 255)),
box_colors))
random.seed(10101) # Fixed seed for consistent colors across runs.
# Shuffle colors to decorrelate adjacent classes.
random.shuffle(box_colors)
# import os
# os.chdir('/Users/xierong/Desktop/Master_2016/googleMLCamp2019/objectDetection/darknet')
import python.darknet as dn
def detection(image, net, meta):
res = dn.detect(net, meta, str.encode(image))
return res
if __name__ == "__main__":
dn.set_gpu(0)
net = dn.load_net(str.encode("cfg/yolov3.cfg"),
str.encode("yolov3.weights"), 0)
meta = dn.load_meta(str.encode("cfg/coco.data"))
image = "data/googleCat.jpg"
res = detection(image, net, meta)
print(res)
# for r in res:
# r = list(r)
# r[0] = str(r[0], encoding = "utf-8")
else:
sys.path.append('.')
import python.darknet as dn
from custom.deep_sort import nn_matching
from custom.deep_sort.tracker import Tracker
from custom.deep_sort.detection import Detection
from custom.application_util import preprocessing
import custom.application_util.geometry as geometry
net_config = {
'weights': b'yolov3.weights',
'config': b'cfg/yolov3.cfg',
'dataset': b'cfg/coco.data',
}
net = dn.load_net(net_config['config'], net_config['weights'], 0)
net_meta = dn.load_meta(net_config['dataset'])
class ObjectTracking:
def __init__(self,
thresh=0.4,
counting_line=None,
nms_max_overlap=1.0,
min_detection_height=20,
max_cosine_distance=0.2,
metric_name='cosine',
classes=('car', 'truck', 'bus')):
self.thresh = thresh
self.counting_line = counting_line
self.nms_max_overlap = nms_max_overlap
# this script must be runned from the darknet folder
input_dir = "../video/"
output_path = "../output.mp4"
data = "cfg/aerial.data"
cfg = "cfg/yolov3-aerial.cfg"
weights = "backup/yolov3-aerial.backup"
images = [d for d in os.listdir(input_dir) if d.endswith('.jpg')]
images.sort()
img = cv2.imread(images[0])
height, width, channels = img.shape
net = dn.load_net(cfg.encode(), weights.encode(), 0)
meta = dn.load_meta(data.encode())
fourcc = cv2.VideoWriter_fourcc(*'mp4v')
out = cv2.VideoWriter(output_path, fourcc, 20.0, (width, height))
count = 0
total = len(images)
for i in images:
count += 1
print("Exportation {}%".format(count / total), end="\r")
img_path = os.path.join(input_dir, i)
r = dn.detect(net, meta, im.encode())
out.write(draw_yolo_result(im, r, ['car', 'bus', 'truck']))
# In[2]:
weights = 'yolov4.weights'
netcfg = 'cfg/yolov4.cfg'
data = 'cfg/coco.data'
# In[ ]:
net = dn.load_net(netcfg.encode('utf-8'), weights.encode('utf-8'), 0)
meta = dn.load_meta(data.encode('utf-8'))
# In[ ]:
cam = 0
# cam = 'rtsp://*****:*****@192.168.88.249:554/Streaming/channels/1'
# cap = camera(cam)
# print(f"Camera is alive?: {cap.p.is_alive()}")
cap = cv2.VideoCapture(cam)
# In[7]: