parser.add_argument('--pretrainEpochs', help='number of epochs for pretraining', default=PRETRAINEPOCHS)
parser.add_argument('--trainEpochs', '-e', help='number of epochs for training', default=TRAINEPOCHS)
parser.add_argument('--batch_size', '-b', help='bastch size', default=BATCH_SIZE)
parser.add_argument('--dropout', help='dropout probability', default=DROPOUT)
parser.add_argument('--optim', help='optimiser', default=OPTIM)
parser.add_argument('--lr', help='learning rate', default=LEARNING_RATE)
parser.add_argument('--cuda', '-g', help='GPU option', default=CUDA, type=bool)
parser.add_argument('--maneuvers', help='maneuvers option', default=MANEUVERS, type=bool)
parser.add_argument('--modelLoc', help='trained prediction store/load location', default=MODELLOC)
parser.add_argument('--pretrain_loss', help='pretrain loss algorithm', default=PRETRAIN_LOSS)
parser.add_argument('--train_loss', help='train loss algorithm', default=TRAIN_LOSS)
parser.add_argument('--list', '-l', action='append', help='DATASet', required=True)
args = parser.parse_args()
model = TnpModel(None)
file_names = []
for i in args.list:
folder = os.path.join(args.dir, DATA_FOLDER.format(i))
video = VIDEO.format(i)
det = 'det.txt'
if args.detection:
sayVerbose(VERBOSE, "begin detection for {}...".format(folder))
model.YOLO_detect(folder, video, args.frames, det, "detectedFrames", args.conf, args.nms, args.cuda)
sayVerbose(VERBOSE, "finished detection for {}...".format(folder))
if args.tracking:
from PyQt5 import QtCore, QtGui, QtWidgets
from PyQt5.QtCore import QThread, pyqtSignal
from view.view import TrackNPredView
from control.controller import Controller
from model.model import TnpModel
if __name__ == "__main__":
import sys
app = QtWidgets.QApplication(sys.argv)
Dialog = QtWidgets.QDialog()
controller = Controller()
view = TrackNPredView()
## to get attributes
view.setupUi(Dialog)
model = TnpModel(controller)
controller.setView(view)
controller.setModel(model)
#show ui
Dialog.show()
sys.exit(app.exec_())
val_lst = apol_to_formatted(val_loc, files, output_dir, "val")
create_data(output_dir, val_lst, args.dir, "val", threadid, class_type)
test_loc = RAW_DATA
output_dir = RAW_DATA + '/test_obs/formatted/'
files = datasets_for_test
test_lst = apol_to_formatted(test_loc, files, output_dir, "test")
create_data(output_dir, test_lst, args.dir, "test", threadid,
class_type)
quit()
print('using {} dataset.'.format(DATASET))
t0 = time.time() #ben: initialize time
model = TnpModel(viewArgs)
if args.cuda:
print("using cuda...\n")
else:
print("using cpu...\n")
if LOAD != '':
model.load(LOAD)
t1 = time.time()
if TRAIN:
model.train(viewArgs['dsId'])
t2 = time.time()
if EVAL:
# output_dir = RAW_DATA + '/val/formatted/'
# files = [f for f in os.listdir(val_loc) if '.csv' in f]
# val_lst = argo_to_formatted(val_loc, output_dir, "val")
# create_data(output_dir, val_lst, args.dir, "val")
# test_loc = RAW_DATA + '/test_obs/data/'
# output_dir = RAW_DATA + '/test_obs/formatted/'
# files = [f for f in os.listdir(test_loc) if '.csv' in f]
# test_lst = argo_to_formatted(test_loc, files, output_dir, "test")
# create_data(output_dir, test_lst, args.dir, "test")
print('using {} dataset.'.format(DATASET))
t0 = time.time()
model = TnpModel(viewArgs)
if args.cuda:
print("using cuda...\n")
else:
print("using cpu...\n")
if LOAD != '':
model.load(LOAD)
t1 = time.time()
# for i in range(5):
# model.train(i)
if TRAIN:
model.train(0)
t2 = time.time()
# ## path settings # args["dir"] = str(self.view.dataDir.text()) # args["frames"] = str(self.view.framesDir.text()) # ## detection settings # args["detection"] = str(self.view.detectionSelector.currentText()) # args["detConf"] = float(self.view.detConfidence.text()) # args["NMS"] = float(self.view.nmsInput.text()) # args["display"] = "False" ## prediction settings args["predAlgo"] = "Traphic" args["pretrainEpochs"] = 6 args["trainEpochs"] = 10 args["batch_size"] = 64 args["dropout"] = .5 args["optim"] = "Adam" args["lr"] = .0001 args["cuda"] = True args["maneuvers"] = False args["modelLoc"] = "resources/trained_models/Traphic_model.tar" args["pretrain_loss"] = '' args['train_loss'] = "MSE" args["dir"] = 'resources/data/TRAF' model = TnpModel() # model.train(args) model.evaluate(args)