import argparse
import cv2
import detect_utils
from PIL import Image
# Construct argument parser
parser = argparse.ArgumentParser()
parser.add_argument('-i', '--input', help='path to input image/video')
parser.add_argument('-m',
'--min-size',
dest='min_size',
default=800,
help='minimum input size for the FasterRCNN network')
args = vars(parser.parse_args())
# Load model
#model = torchvision.models.detection.fasterrcnn_resnet50_fpn(pretrained=True,
#min_size=args['min_size'])
model = torch.load("model_epoch18.pth", map_location=torch.device('cpu'))
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
image = Image.open(args['input'])
model.eval().to(device)
boxes, classes, labels = detect_utils.predict(image, model, device, 0.8)
image = detect_utils.draw_boxes(boxes, classes, labels, image)
cv2.imshow('Image', image)
save_name = "Pred_img"
cv2.imwrite(f"outputs/{save_name}.jpg", image)
cv2.waitKey(0)
[{
"image_id": int, "category_id": int, "bbox": [x,y,width,height], "score": float,
}]
"""
annotations = []
# read until end of video
while (cap.isOpened()):
# capture each frame of the video
ret, frame = cap.read()
if ret == True:
# get the start time
start_time = time.time()
with torch.no_grad():
# get predictions for the current frame
boxes, classes, labels, scores = detect_utils.predict(
frame, model, device, 0.0)
#generate json objects in correct form for each frame
# print(boxes)
# print(classes)
# print(labels)
# print(scores)
for i in range(len(boxes)):
print(labels[i].item())
print(boxes[i].tolist())
print(scores[i])
data = {}
data['image_id'] = int(frame_count)
data['category_id'] = int(labels[i].item())
data["bbox"] = boxes[i].tolist()
data["score"] = float(scores[i])
frame_height = int(cap.get(4)) #mendapatkan tinggi frame
save_name = f"{args['input'].split('/')[-1].split('.')[0]}_{args['min_size']}" #saving input
out = cv2.VideoWriter(
f"outputs/{save_name}.mp4", cv2.VideoWriter_fourcc(*'mp4v'), 30,
(frame_width, frame_height)) # define codec and create VideoWriter object
frame_count = 0 # menghitung jumlah frame
total_fps = 0 # mendapatkan final FPS
while (cap.isOpened()): # membaca vidoe sampai berakhir
ret, frame = cap.read() # capture tiap frame pada video
if ret == True:
start_time = time.time() #start time
with torch.no_grad():
boxes, classes, labels = detect_utils.predict(
frame, model, device,
0.8) # get predictions for the current frame
image = detect_utils.draw_boxes(
boxes, classes, labels,
frame) # draw boxes and show current frame on screen
counter = detect_utils.Counter(classes,
nama_kelas='car') # counter class
end_time = time.time() # get the end time
fps = 1 / (end_time - start_time) # get the end time
total_fps += fps # get the end time
frame_count += 1 # increment frame count
wait_time = max(1, int(fps / 4))
cv2.imshow('image', image)
out.write(image)
if cv2.waitKey(wait_time) & 0xFF == ord('q'):
parser.add_argument('-t', '--threshold', default=0.6, type=float,
help='minimum confidence score for detection')
args = vars(parser.parse_args())
print('USING: ')
print(f"Minimum image size: {args['min_size']}")
print(f"Confidence threshold: {args['threshold']}")
# download or load the model from disk
model = torchvision.models.detection.retinanet_resnet50_fpn(pretrained=True,
min_size=args['min_size'])
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
# load the model onto the computation device
model.eval().to(device)
image = Image.open(args['input']).convert('RGB')
# a Numpy copy for OpenCV functions
image_array = np.array(image)
# convert to OpenCV BGR to color format
image_array = cv2.cvtColor(image_array, cv2.COLOR_RGB2BGR)
# get the bounding boxes and class labels
boxes, classes = detect_utils.predict(image, model, device, args['threshold'])
# get the final image
result = detect_utils.draw_boxes(boxes, classes, image_array)
cv2.imshow('Image', result)
cv2.waitKey(0)
save_name = f"{args['input'].split('/')[-1].split('.')[0]}_{args['min_size']}_t{int(args['threshold']*100)}"
cv2.imwrite(f"outputs/{save_name}.jpg", result)
frame_count = 0 # to count total frames
total_fps = 0 # to get the final frames per second
# load the model onto the computation device
model = model.eval().to(device)
# read until end of video
while (cap.isOpened()):
# capture each frame of the video
ret, frame = cap.read()
if ret == True:
# get the start time
start_time = time.time()
with torch.no_grad():
# get predictions for the current frame
boxes, classes, labels = predict(frame, model, device, 0.8)
# draw boxes and show current frame on screen
image = draw_boxes(boxes, classes, labels, frame)
# get the end time
end_time = time.time()
# get the fps
fps = 1 / (end_time - start_time)
# add fps to total fps
total_fps += fps
# increment frame count
frame_count += 1
# press `q` to exit
wait_time = max(1, int(fps / 4))
cv2.imshow('image', image)
out.write(image)
