def save_model(model, tort = 'Val'):
images_so_far = 0
if tort == 'Train':
dataloader = dataloader_train
else:
dataloader = dataloader_val
for i, data in enumerate(dataloader):
samples = data
if use_gpu:
inputs = Variable(samples['image'].cuda())
trajectories = Variable(samples['trajectories'].cuda())
else:
inputs, trajectories = Variable(samples['image']), Variable(samples['trajectories'])
outputs = model(inputs)
for j in range(inputs.size()[0]):
outData = GMR(outputs[j], trajectories[j])
images_so_far += 1
file_path = 'predictedTraj' + tort + '/' + str(images_so_far) + '.txt'
np.savetxt(file_path, outData.cpu().data.numpy())
def trajPredictor(model, image):
mean = np.array([0.485, 0.456, 0.406])
std = np.array([0.229, 0.224, 0.225])
transform_data = transforms.Compose([
Rescale((240, 320)),
#Normalize(mean, std),
ToTensor()
])
if use_gpu:
inputs = Variable(transform_data(image).cuda())
else:
inputs = Variable(transform_data(image))
inputs = inputs.unsqueeze(0)
outputs = model(inputs)
traj = GMR(outputs[0])
if use_gpu:
nptraj = traj.data.cpu().numpy()
else:
nptraj = traj.data.numpy()
return nptraj
def visualize_model(model, num_images=4, tort = 'Val'):
images_so_far = 0
fig = plt.figure()
model.train(False)
traj_len = 200
if use_gpu:
torchType = torch.cuda.FloatTensor
else:
torchType = torch.FloatTensor
time = Variable(torch.arange(0, (traj_len - 1) / traj_len, 1 / traj_len).type(torchType) , requires_grad=False)
if tort == 'Train':
dataloader = dataloader_train
else:
dataloader = dataloader_val
for i, data in enumerate(dataloader):
samples = data
if use_gpu:
inputs = Variable(samples['image'].cuda())
trajectories = Variable(samples['trajectories'].cuda())
else:
inputs, trajectories = Variable(samples['image']), Variable(samples['trajectories'])
outputs = model(inputs)
for j in range(outputs.size()[0]):
outData = GMR(outputs[j])
loss = (trajectories[j] - outData).pow(2).sum().sqrt() / 200
print(loss)
images_so_far += 1
ax = plt.subplot(num_images//2, 2, images_so_far)
ax.axis('off')
inp = inputs[j].cpu().data.numpy().transpose((1, 2, 0))
#mean = np.array([0.485, 0.456, 0.406])
#std = np.array([0.229, 0.224, 0.225])
#inp = std * inp + mean
#inp = np.clip(inp, 0, 1)
ax.imshow(inp, origin='upper')
img_traj = trajectories[j].cpu().data.numpy()
img_out_data = outData.cpu().data.numpy()
# img_traj[:, 1] = (img_traj[:, 1] + 0.75) * 240
# img_traj[:, 0] = (img_traj[:, 0] + 1.0) * 240
# img_out_data[:, 1] = (img_out_data[:, 1] + 0.75) * 240
# img_out_data[:, 0] = (img_out_data[:, 0] + 1.0) * 240
img_traj[:, 0] = (img_traj[:, 0] + 1.0) * 240
img_traj[:, 1] = (img_traj[:, 1] + (2.0 / 3.0)) * 240
img_out_data[:, 0] = (img_out_data[:, 0] + 1.0) * 240
img_out_data[:, 1] = (img_out_data[:, 1] + (2.0 / 3.0)) * 240
#imshow(inputs.cpu().data[j])
plt.scatter(img_traj[:, 1], img_traj[:, 0], s=10, marker='.', color='red')
plt.scatter(img_out_data[:, 1], img_out_data[:, 0], s=10, marker='.', color='blue')
# plt.scatter(time.cpu().data.numpy(), trajectories[j][:, 0].cpu().data.numpy(), s=10, marker='.', color='red')
# plt.scatter(time.cpu().data.numpy(), outData[:, 0].cpu().data.numpy(), s=10, marker='.', color='blue')
# plot_results(outputs[j], ax)
# plot_results_time(outputs[j], ax)
plt.pause(0.001)
if images_so_far == num_images:
return
image_torch = torch.from_numpy(image).float()
if use_gpu:
inputs = Variable(image_torch.cuda())
else:
inputs = Variable(image_torch)
inputs = inputs.unsqueeze(0)
outputs = model(inputs)
fig = plt.figure()
nameFile = save_dir + str(i) + '.txt'
np.savetxt(nameFile, outputs[0].cpu().data.numpy())
outData = GMR(outputs[0])
inp = inputs[0].cpu().data.numpy().transpose((1, 2, 0))
plt.imshow(inp, origin='upper')
plt.axis('off')
img_out_data = outData.cpu().data.numpy()
img_out_data[:, 0] = (img_out_data[:, 0] + 1.0) * 240
img_out_data[:, 1] = (img_out_data[:, 1] + (2 / 3)) * 240
plt.scatter(img_out_data[:, 1],
img_out_data[:, 0],
s=10,
marker='.',
color='blue')
def __init__(self):
self.gmr = GMR()
self.launch_gmr()
class SHO_APP:
def __init__(self):
self.gmr = GMR()
self.launch_gmr()
def set_gmr_sleep_delay(self, sleep_delay):
self.gmr.set_sleep_delay(sleep_delay)
def set_gmr_time_for_push_notification(self, from_, to):
self.gmr.set_time_for_push_notification(from_, to)
def launch_gmr(self):
self.gmr.set_mode(True)
self.gmr.launch()
def set_light_delay(self, delay):
payload = {
"datapoints": [{
"dataChnId": "4",
"values": {
"value": delay
}
}]
}
self.gmr.post_to_mcs(payload)
def set_automation_button(self, value):
payload = {
"datapoints": [{
"dataChnId": "3",
"values": {
"value": value
}
}]
}
self.gmr.post_to_mcs(payload)
def set_control_button(self, value):
payload = {
"datapoints": [{
"dataChnId": "2",
"values": {
"value": value
}
}]
}
self.gmr.post_to_mcs(payload)
def get_light_delay(self):
return int(self.gmr.get_from_mcs("4"))
def get_automation_button(self):
return int(self.gmr.get_from_mcs("3"))
def get_control_button(self):
return int(self.gmr.get_from_mcs("2"))