import cv2
import numpy as np
from rknn.api import RKNN
if __name__ == '__main__':
rknn = RKNN(verbose=False)
rknn.register_op('./resize_area/ResizeArea.rknnop')
rknn.load_tensorflow(tf_pb='./resize_area_test.pb',
inputs=['input'],
outputs=['resize_area_0'],
input_size_list=[[32, 32, 3]])
rknn.build(do_quantization=False)
# rknn.export_rknn('./resize_area.rknn')
# rknn.load_rknn('./resize_area.rknn')
rknn.init_runtime()
img = cv2.imread('./dog_32x32.jpg')
outs = rknn.inference(inputs=[img])
out_img = outs[0].astype('uint8')
out_img = np.reshape(out_img, (64, 64, 3))
cv2.imwrite('./out.jpg', out_img)
orig_rknn = sys.argv[1]
hw_rknn = sys.argv[2]
# Create RKNN object
rknn = RKNN()
# Load rknn model
print('--> Loading RKNN model')
ret = rknn.load_rknn(orig_rknn)
if ret != 0:
print('Load RKNN model failed!')
exit(ret)
print('done')
# Init runtime environment
print('--> Init runtime environment')
# Note: you must set rknn2precompile=True when call rknn.init_runtime()
# RK3399Pro with android system does not support this function.
ret = rknn.init_runtime(target='rk1808', rknn2precompile=True)
if ret != 0:
print('Init runtime environment failed')
exit(ret)
print('done')
ret = rknn.export_rknn_precompile_model(hw_rknn)
rknn.release()
print('--> Export RKNN model')
ret = rknn.export_rknn('./xception.rknn')
if ret != 0:
print('Export xception.rknn failed!')
exit(ret)
print('done')
# ret = rknn.load_rknn('./xception.rknn')
# Set inputs
img = cv2.imread(IMG_PATH)
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
# init runtime environment
print('--> Init runtime environment')
ret = rknn.init_runtime()
# ret = rknn.init_runtime(target='rk1808')
if ret != 0:
print('Init runtime environment failed')
exit(ret)
print('done')
# Inference
print('--> Running model')
outputs = rknn.inference(inputs=[img])
show_outputs(outputs)
print('done')
rknn.release()
def run_ssd(img_path,priorbox_path):
#caffe_proto="./MobileNetSSD_deploy.prototxt"
caffe_proto= "./MobileNetSSD_deploy_truncated.prototxt"
caffe_weight="./MobileNetSSD_deploy10695.caffemodel"
rknn_model="./pedestrian_ssd.rknn"
caffe2rknn(caffe_proto,caffe_weight,rknn_model)
print("run ssd")
rknn=RKNN(verbose=True)
ret=rknn.load_rknn(path=rknn_model)
ret=rknn.init_runtime()
#ret = rknn.init_runtime(target='rk1808', device_id='012345789AB')
img=cv2.imread(img_path)
img=cv2.resize(img,(300,300))
print("shape:",img.shape)
outlen=7668 #change to your model
priorbox=[]
with open(priorbox_path) as f:
for line in f:
arr=line.strip().split(",")
priorbox=list(map(float,arr))
priorbox=np.reshape(np.array(priorbox),(2,outlen))
outputs = rknn.inference(inputs=[img])#,data_format="nchw",data_type="float32"
print("pb:",priorbox.shape,priorbox)
print("loc:",outputs[0].shape,outputs[0])
print("conf:",outputs[1].shape,outputs[1])
NUM_RESULTS=outlen//4
NUM_CLASSES=2
box_priors= priorbox[0].reshape((NUM_RESULTS,4))
box_var = priorbox[1].reshape((NUM_RESULTS,4))
loc = outputs[0].reshape((NUM_RESULTS, 4))
conf = outputs[1].reshape((NUM_RESULTS, NUM_CLASSES))
#compute softmax
conf = [[x/(x+y),y/(x+y)] for x,y in np.exp(conf)]
# Post Process
for i in range(0, NUM_RESULTS):
pb = box_priors[i]
lc = loc[i]
var= box_var[i]
pb_w = pb[2] - pb[0]
pb_h = pb[3] - pb[1]
pb_cx = (pb[0] + pb[2]) * 0.5;
pb_cy = (pb[1] + pb[3]) * 0.5;
bbox_cx = var[0] * lc[0] * pb_w + pb_cx;
bbox_cy = var[1] * lc[1] * pb_h + pb_cy;
bbox_w = math.exp(var[2] * lc[2]) * pb_w;
bbox_h = math.exp(var[3] * lc[3]) * pb_h;
xmin = bbox_cx - bbox_w * 0.5;
ymin = bbox_cy - bbox_h * 0.5;
xmax = bbox_cx + bbox_w * 0.5;
ymax = bbox_cy + bbox_h * 0.5;
xmin *= 300 #input width
ymin *= 300 #input height
xmax *= 300 #input width
ymax *= 300 #input height
score = conf[i][1];
if score > 0.9:
print("score:",score)
cv2.rectangle(img, (int(xmin), int(ymin)), (int(xmax), int(ymax)),(0, 0, 255), 3)
plt.imshow(cv2.cvtColor(img,cv2.COLOR_RGB2BGR))
plt.show()
print("ssd finished")
rknn.build(do_quantization=True, dataset='./dataset.txt')
print('done')
# Export RKNN Model
rknn.export_rknn('./ssd_mobilenet_v1_coco.rknn')
# Set inputs
orig_img = cv2.imread('./road.bmp')
img = cv2.cvtColor(orig_img, cv2.COLOR_BGR2RGB)
img = cv2.resize(img, (INPUT_SIZE, INPUT_SIZE),
interpolation=cv2.INTER_CUBIC)
# init runtime environment
print('--> Init runtime environment')
# ret = rknn.init_runtime()
ret = rknn.init_runtime(target='rk1808')
if ret != 0:
print('Init runtime environment failed')
exit(ret)
print('done')
times = 0
first_outputs = None
while True:
# Inference
times += 1
print('--> Running model: %d' % (times))
outputs = rknn.inference(inputs=[img])
# print('', type(outputs[0]), type(outputs[1]))
print('inference result: ', outputs)
if not first_outputs:
def main(folder="test"):
folder = folder
files = os.listdir(folder)
for i in range(len(files)):
img = cv2.imread("{}/{}".format(folder, files[i]))
img = (img - 127.5) / 127.5
h, w = img.shape[:2]
print("w, h = ", w, h)
input = cv2.resize(img, (PRESET, PRESET),
interpolation=cv2.INTER_CUBIC)
input = input.reshape(PRESET, PRESET, 3)
input = np.array(input, dtype=np.float32)
rknn = RKNN()
print('--> Loading model')
#rknn.config(channel_mean_value='0 0 0 255', reorder_channel='0 1 2')
# Load TensorFlow Model
print('--> Loading model')
rknn.load_tensorflow(tf_pb='pretrained/SR_freeze.pb',
inputs=['ESRGAN_g/Conv2D'],
outputs=['output_image'],
input_size_list=[[PRESET, PRESET, 3]])
print('done')
# Build Model
print('--> Building model')
rknn.build(do_quantization=False)
print('done')
# Export RKNN Model
rknn.export_rknn('./sr_rknn.rknn')
# Direct Load RKNN Model
rknn.load_rknn('./sr_rknn.rknn')
# init runtime environment
print('--> Init runtime environment')
ret = rknn.init_runtime()
if ret != 0:
print('Init runtime environment failed')
# Inference
print('--> Running model')
output_image = rknn.inference(inputs=[input])
print('complete')
out = np.array(output_image, dtype=np.float64)
print("output_image = ", out.shape)
out = np.squeeze(out)
Y_ = out.reshape(PRESET * 4, PRESET * 4, 3)
Y_ = cv2.resize(Y_, (w * 4, h * 4), interpolation=cv2.INTER_CUBIC)
print("output shape is ", Y_.shape)
#후처리 과정
Y_ = (Y_ + 1) * 127.5
cv2.imwrite("{}/{}_yval.png".format(OUT_DIR, i), Y_)
# Evaluate Perf on Simulator
#rknn.eval_perf()
# Release RKNN Context
rknn.release()
def run(self): #把要执行的代码写到run函数里面 线程在创建后会直接运行run函数
minsize = 20
threshold = [0.8, 0.9, 0.95]
factor = 0.709
pnet_rknn_list = init_pnet()
rnet_rknn = RKNN()
onet_rknn = RKNN()
rnet_rknn.load_rknn('./RNet.rknn')
onet_rknn.load_rknn('./ONet.rknn')
ret = rnet_rknn.init_runtime()
if ret != 0:
#print('Init rnet runtime environment failed')
exit(ret)
ret = onet_rknn.init_runtime()
if ret != 0:
#print('Init onet runtime environment failed')
exit(ret)
sys.stdout = open('/dev/stdout', 'w')
sys.stderr = open('/dev/stderr', 'w')
global proflag
global IMAGE_list
global boundingbox_list
nonfacecount = 0
#wrongimg = 1
while (proflag == 1):
imreadLock.acquire()
img0 = IMAGE_list[0].copy()
img = IMAGE_list[1].copy()
imreadLock.release()
#tic()
score_cmp = compare_image(img0, img)
#print("score_cmp",score_cmp)
#toc()
if score_cmp < 0.98:
#imreadLock.release()
#print("detect face start")
#cv2.imwrite("aa.jpg",img)
tic()
boundingboxes, points = detect_face(img, minsize,
pnet_rknn_list, rnet_rknn,
onet_rknn, threshold,
False, factor)
#print("boundingboxes shape",boundingboxes.shape)
print("total cost")
toc()
if boundingboxes.shape[0] != 0:
if len(boundingbox_list) != 0:
boundingbox_list.clear()
boundingbox_list.append(boundingboxes)
else:
#path = str(wrongimg)+".jpg"
#cv2.imwrite(path,img)
#wrongimg += 1
nonfacecount += 1
if nonfacecount >= 3:
boundingbox_list.clear()
nonfacecount = 0
for i in range(2):
pnet_rknn_list[i].release()
rnet_rknn.release()
onet_rknn.release()
name = name[:name.find('.')]
eval_img_name.append(name)
eval_low_im = load_images(eval_low_data_name[idx])
eval_low_data.append(eval_low_im)
print(eval_low_im.shape) # (400, 600, 3)
height, width, channels = eval_low_im.shape # test 사진의 원본 크기 저장 >> 최종이미지 저장시 사용함
input_low = eval_low_data[idx]
input_low_eval = np.expand_dims(input_low, axis=0) # (1, 400, 600, 3)
sample_dir = './results/test/'
if not os.path.isdir(sample_dir):
os.makedirs(sample_dir)
# DecomNet 실행
print('-> Running model 1')
ret1 = rknn.init_runtime()
if ret1 != 0:
print('Init runtime environment 1 failed')
exit(ret1)
print('done')
print('-> Inference model 1')
decom_r_low, decom_i_low = rknn.inference(inputs=[input_low_eval]) # decom_r_low(3채널)
#print(type(decom_r_low)) #numpy.ndarray
print('=> 1 run success')
#image save (성공)
save_images(os.path.join(sample_dir, 'decom_r_low.png'), decom_r_low)
save_images(os.path.join(sample_dir, 'decom_i_low.png'), decom_i_low)
#------------------------------------------------------------------------------------------------------------
