async def train(ctx, filename: str, memsize: int):
try:
with open("training_data/" + filename + ".txt", "r") as f:
tdata = f.read()
except IOError:
await ctx.send("Error - Specified model does not exist.")
return
model = ai.train(tdata, memsize)
with open("models/" + filename + "-" + str(memsize) + ".model", "w") as f:
f.write(json.dumps([model, tdata]))
await ctx.send("Done!")
parser_train = subparsers.add_parser(
'train',
help=
'trains the system to detect traffic lights in an image given training samples'
)
# command: predict; for predicting traffic lights given an input image
parser_predict = subparsers.add_parser(
'predict', help='attempts to find all traffic lights in an image')
parser_predict.add_argument('--path', help='path to image', required=True)
# command: mine; for mining positive and negative examples from a given input image
parser_mine = subparsers.add_parser(
'mine',
help='tool for hard mining positive and negative training data')
parser_mine.add_argument('--path', help='path to image', required=True)
parser_mine.add_argument(
'--use-predicted',
help=
'whether or not to use the boxes that were predicted as positive samples',
action='store_true')
args = parser.parse_args()
if args.subcmd == 'train':
ai.train()
elif args.subcmd == 'predict':
ai.predict(path.expanduser(args.path))
elif args.subcmd == 'mine':
ai.mine(path.expanduser(args.path), args.use_predicted)
cTile = tileController.getTile( index )
index += 1
nTile = tileController.getTile( index )
viewController.setTile( cTile, nTile )
viewController.updateEverything( )
"""
for j in range(5):
times = 0
index = 0
ai.totalReward = 0
for i in range(10000):
if timeController.timeEvent():
if viewController.aiState:
#move, rotate, rating = ai.makeMove( cTile )
ai.train(cTile)
if not cTile.incY():
cTile.apply()
if not gridController.checkForGameOver():
scoreController.tileReleased()
cTile = nTile
index += 1
nTile = tileController.getTile(index)
viewController.setTile(cTile, nTile)
else:
times += 1
cTile = tileController.getTile(index)
index += 1
nTile = tileController.getTile(index)
viewController.setTile(cTile, nTile)
val = arg.split("=")[1]
if data.get(key) != None:
data[key] = val
else:
usage()
else:
usage()
if sys.argv[1] == "train":
if data["training-file"] == "stdin":
tdata = input()
else:
with open(data["training-file"], "r") as f:
tdata = f.read()
model = ai.train(tdata, int(data["max-history"]))
if data["model-file"] == "stdout":
print(json.dumps(model))
else:
with open(data["model-file"], "w") as f:
f.write(json.dumps([model, tdata]))
elif sys.argv[1] == "predict":
if data["model-file"] == "stdin":
model = input()
else:
with open(data["model-file"], "r") as f:
model = json.loads(f.read())[0]
size = 8
from io import BytesIO
from PIL import Image
import numpy as np
import requests
import ai
ai_ = None
flag = True
while True:
url = 'http://0.0.0.0:5000/image/verification'
res = requests.get(url)
image = Image.open(BytesIO(res.content))
image.show()
length, width = image.size
if flag:
user = input('your chose:\n')
if user == 'ai':
url = 'http://0.0.0.0:5000/ai/{}'.format(ai_)
flag = False
res = requests.get(url)
print(res.text)
else:
check_list = [[int(temp) for temp in user]]
input_data = np.array(image)
np.reshape(input_data, (length, width, 3))
print(check_list)
ai_ = ai.train(length, width, [input_data], check_list)
print(ai_)