did you finish the repo or not yet, causing I got error.

[vxgu86]

Loading existing training The RGB camera returned bad data so Im creating zero array to not break the training Traceback (most recent call last):

File "", line 1, in runfile('D:/RL/4DroneTracking-master/DQN-Train.py', wdir='D:/RL/DroneTracking-master')

File "D:\Anaconda3\lib\site-packages\spyder_kernels\customize\spydercustomize.py", line 668, in runfile execfile(filename, namespace)

File "D:\Anaconda3\lib\site-packages\spyder_kernels\customize\spydercustomize.py", line 108, in execfile exec(compile(f.read(), filename, 'exec'), namespace)

File "D:/RL/4DroneTracking-master/DQN-Train.py", line 155, in dqn.fit(env, callbacks=callbacks, nb_steps=250002, visualize=False, verbose=0, log_interval=5000)

File "D:\Anaconda3\lib\site-packages\rl\core.py", line 169, in fit action = self.forward(observation)

File "D:\Anaconda3\lib\site-packages\rl\agents\dqn.py", line 228, in forward q_values = self.compute_q_values(state)

File "D:\Anaconda3\lib\site-packages\rl\agents\dqn.py", line 69, in compute_q_values q_values = self.compute_batch_q_values([state]).flatten()

File "D:\Anaconda3\lib\site-packages\rl\agents\dqn.py", line 64, in compute_batch_q_values q_values = self.model.predict_on_batch(batch)

File "D:\Anaconda3\lib\site-packages\keras\engine\training.py", line 1268, in predict_onbatch x, , _ = self._standardize_user_data(x)

File "D:\Anaconda3\lib\site-packages\keras\engine\training.py", line 751, in _standardize_user_data exception_prefix='input')

File "D:\Anaconda3\lib\site-packages\keras\engine\training_utils.py", line 128, in standardize_input_data 'with shape ' + str(data_shape))

ValueError: Error when checking input: expected input_1 to have 4 dimensions, but got array with shape (1, 1, 90, 256, 3)

///////////////////////////////////////////////////////////////////////// in getScreenSceneVis()
img1d = np.array(bytearray(responses[0].image_data_uint8))#.reshape(responses[0].height, responses[0].width, 4) if len(img1d) != (1442564): ----->should this be 1442563?

I find the same code in your TrackGym also, what's the meaning?

[vxgu86]

@AloshkaD sorry to bother

[AloshkaD]

This repo is deprecated and I'm going to push a new one with major upgrades. It's currently testing on my machines, so I can't tell exactly when will it be ready. Hopfully in a month.

[vxgu86]

sorry for the bother the problem should be easy for you, and I am working on this right now We are getting a rgb image, it should be 1442563, why it's 4??

getScreenSceneVis() img1d = np.array(bytearray(responses[0].image_data_uint8))#.reshape(responses[0].height, responses[0].width, 4) if len(img1d) != (1442564): ----->should this be 1442563?

[vxgu86]

and one question, are you using the visual studio as IDE?

[AloshkaD]

No, we are not getting RGB from airsim. The 4 channels are for RGB and gamma. I'm using Microsoft code as I develop on Linux.

[vxgu86]

sorry but could you explain the detail of getScreenSceneVis, you are using simGetImages, it's from AirSim. by which way can I run your code up?

[AloshkaD]

no problem. I'm a bit busy these days as I'm preparing for graduation. I'll do my best to answer the questions that haven't been addressed on Airsim's repo. Since this method is from airsim, I highly recommend that you read airsim's api and the docs. There you can find detailed explanation for how to call rgb or depth images.

[vxgu86]

sorry..... but I have red through the api docs include https://github.com/microsoft/AirSim/blob/master/docs/image_apis.md without finding anythin about the 4 channels of RGB and gamma

[AloshkaD]

I'm sorry. I meant to say Alph and not Gamma. My bad.So It's RGBA. Check out this https://github.com/microsoft/AirSim/issues/1557

[vxgu86]

this is great, thanks but my return of img1d = np.array(bytearray(responses[0].image_data_uint8)) len(img1d) equals 144x256x3, why it's not 4 channels?

should I reshape it like in https://github.com/microsoft/AirSim/issues/1557 as follows: img4d_rgba = img1d_rgba.reshape(512, 512, 4) img3d_rbg = img4d_rgba[:, :, :3]

[AloshkaD]

Best thing to do is to plot the image and check what you are getting. You need to reshape the image but depending on the channels you either do 1, 3 or 4