About saving the image to disc: As far as I know, training is more efficient when you randomly draw from memory samples. So saving a large amount of samples (together with rewards of course which is the harder part) makes sense
@MrConi10
5 жыл бұрын
Are you speaking about building an experience replay "buffer" ? That could be true in some cases (if you have a lot of FPS and a small buffer -- leading to nearly no changes between all frames) but dangerous on some others. For example it could lead to issues as you don't have the past states / actions. I'm also wondering if DQN is a working approach for self-driving car ?
@machwave540
5 жыл бұрын
Feel sorry for those people who fell down while watching the video and accidentally pressed the dislike button. Get well soon guys.
@asdfasdf-vp9up
5 жыл бұрын
LMAO
@dashawnlyons2791
4 жыл бұрын
LMAOOOOOOOOO
@mohammedalsubaie3512
4 жыл бұрын
THE LOCALHOST has changed in the new version of Carla. this is the replacement ('127.0.0.1') I don't really know if it has a meaning but this is what I found. if you can explan=in what does it mean I would really appreciate that. and really thank you so much for those amazing tutorials.
@eudo81
5 жыл бұрын
Part 2... and the party is on. Good job sentdex!
@nyctophilic1790
4 жыл бұрын
Hi Sentdex thanks for the tutorial, i think save_to_disk() method is very helpful if you want to collect a dataset to train neural networks beside RL for object detection, segmentation or behavioural cloning.
@marcitakacs1271
5 жыл бұрын
Nice video as Always sentdex.
@BublGum3
3 жыл бұрын
Was that Runelite flashed on the screen for a second. OSRS, nice haha
@orhunonar
4 ай бұрын
Hi Sentdex. How can I find the actual X and Z point on the vehicle. Do I need a Blender for this ?
@saad-ul4mr
5 жыл бұрын
You can do data collection for self driving cars 🙌
@xXKM4UXx
3 жыл бұрын
Is anyone else getting a blackscreen when running the python module?
@vickyskarthik7022
3 жыл бұрын
Me too. Did you find the solution for it?
@whitekyurem1
5 жыл бұрын
Hey sentdex! Amazing videos. I was wondering whether its possible to train a model using supervised learning from real life video data, and then test it out on the Carla simulator.
@shaanlashkari8769
3 жыл бұрын
Great Question , even i have the same doubt!
@saravanajogan1221
3 жыл бұрын
Sir, please give me ideas on how can create my own roads with potholes and train the vehicle to slow down when it sees a pothole...Iam trying sir but I don't know how to start...
@emilyzhang2099
4 жыл бұрын
Hi, you run the python code very fast, and the carlaUE4 is very fluent. But on my computer, the CPU is 100% occupied, and GPU is about 50% occupied when the py code is run. My computer configuration is 64bit, Core i5, 3GHz, and the RAM is 16G, GeForce RTX 2060 14G. The CPU is 100% occupied, so it will crash after several-second running. How about your computer configuration?
@prashantkramadhari3724
5 жыл бұрын
Hey sentdex, Which algorithm from RL you plan to try...?Is it too much if i ask whats the end application(driving fast without collision and following rules)..?
@sentdex
5 жыл бұрын
I don't really know. My goal is what I said: Drive fast, avoid crashing. That might change, we'll see :D
@bobmishra3698
5 жыл бұрын
Hey man whoever is reading is remember you are awesome and don't let others judgement bring you down. Have a Awesome day!
@Aish5176
5 жыл бұрын
Hey, i have a problem while running the code on my ubuntu system. i was not able to get the camera view and it says that it was able to connect to X Server , can you please help me out?
@hafsamsaaf232
4 жыл бұрын
i want to ask how can i resolve this error attribute eror: module carla has no attribute "client"
@yolandataniamulyadi4456
3 жыл бұрын
Can anyone help me? I can't import cv2 even though I've already installed opencv-python. Is it because I have python 3.8?? If I run "python manual_control_edited.py", line of "import cv" not showing error but there is error in "import carla". But if I run "py -3.7 manual_control_edited.py", line of "import cv"showing error
@lennartlut
5 жыл бұрын
32:04 I followed your tutorial. I have difficulties displaying the correct cv2.imshow . It's completly white and I couldn't figure out how it is working properly for you, but not for me. I was able to write all the images into a list and display the images in the "main" thread afterwards with cv2.imshow. So it appears to be working in general, just you implementation doesn't work for me.
@lennartlut
5 жыл бұрын
def process_img(image, l_images): i = np.array(image.raw_data) # print(i.shape) i2 = i.reshape((IM_HEIGHT, IM_WIDTH, 4)) #rgba, a for alpha i3 = i2[:, :, :3] # /255.0 # entire height, entire width, only rgb (no alpha) print(i3[1 , 1, :]) #import pdb; pdb.set_trace() #cv2.imshow("image", i3) #cv2.waitKey(0) l_images.append(i3) return #i3/255.0 # normalize the data [...] l_images = [] sensor.listen(lambda data: process_img(data, l_images)) time.sleep(1) sensor.stop() for im in l_images: cv2.imshow('image', im) cv2.waitKey(16)
@@lennartlut I was having the same issue, this solution worked for me.
@yutingwu9260
2 жыл бұрын
your solution helps
@ankurkumarsrivastava6958
4 жыл бұрын
cv2.imshow("", i3) shows only a white screen, anybody else facing similar issue with 0.9.9 carla?
@zhohtroboht3540
4 жыл бұрын
I'm facing the same issue with Carla 0.9.9. Had anyone come up with a solution?
@zhohtroboht3540
4 жыл бұрын
It appears to be the problem with OpenCV, not Carla.
@mehdiozel517
4 жыл бұрын
Me havin' same problem. Probably, because of Opencv version
@mehdiozel517
4 жыл бұрын
I solved the problem. Make it -> cv2.waitKey(0)
@13579Josiah
3 жыл бұрын
I'm having this problem, using CARLA 0.9.6 and Ubuntu 20.10.0. cv2.waitKey(0) doesn't work for me.
@abhishekchandra5927
5 жыл бұрын
Hello Sentdex I have a question please answer I want to make software on python are there any online platform for coding, deploy basically i want to build an online application which is able to run without my computer just like a website? Please help me i'm very Confused😕😕😕😕
@TnInventor
5 жыл бұрын
like first , watch later
@louahem1
4 жыл бұрын
Greetings! To those who received this error: Nandkumar Sanodiya 8 months ago File "tutorial-video.py", line 38, in world = client.get_world() RuntimeError: time-out of 2000ms while waiting for the simulator, make sure the simulator is ready and connected to localhost:2000 I tried this solution and it worked: #first add this library to import import argparse #second add this method before the "try:" section: argparser = argparse.ArgumentParser( description=__doc__) argparser.add_argument( '--host', metavar='H', default='127.0.0.1', help='IP of the host server (default: 127.0.0.1)') argparser.add_argument( '-p', '--port', metavar='P', default=2000, type=int, help='TCP port to listen to (default: 2000)') args = argparser.parse_args() #call the client using the args: try: client = carla.Client(args.host, args.port) client.set_timeout(10.0) I hope it helps, I don't know why it works really I just compared with other doc code and did some pragmatic trial and error.
@aymen_bhd
4 жыл бұрын
Still the same issue brother
@korean4130
4 жыл бұрын
I don't know what it did, but it worked :) Thanks!
@aydinaltun3816
3 жыл бұрын
if anyone encournter with RuntimeError problem just change the time in "client.set_timeout(2.0)" in the code to "client.set_timeout(200.0)" # 200 depicts the loading time.
@matthewao
5 жыл бұрын
When I run the script after having copied everything, I get this: ActorBlueprint(id=vehicle.tesla.model3,tags=[model3, tesla, vehicle]) All cleaned up! LIFETIME: WARNING! the following objects were not destructed. LIFETIME: tcp client 2 still alive. LIFETIME: Actor 47 (sensor.camera.rgb) still alive. LIFETIME: Actor 46 (vehicle.tesla.model3) still alive. LIFETIME: tcp client 1 still alive. LIFETIME: SimulatorClient(localhost:2000) still alive. LIFETIME: AsioThreadPool still alive. What's going wrong here?
@omegaaq2329
Жыл бұрын
Now I realized that my knowledge is complete shit. I😅 can forget about Men's career as a programmer
@aliusa6327
5 жыл бұрын
Nice job sentdex you are the Python expert 👍keep up bro
@Craskar
4 жыл бұрын
Friends, I am having trouble with connecting to the localhost. Doing it once is fine and I can execute the python skripts, no problem. But if I close and reopen carla, rerun the skripts, it doesn't connect. Anyone haveing the same issue?
@dyangzng6388
3 жыл бұрын
I met the same problem.do you know how to solve it?
@h4ndle_yt
5 жыл бұрын
Being a Microsoft oriented guy, I'm surprised you didn't go for AirSim instead of CARLA.
@Watashiwapitadesu
3 жыл бұрын
Hey great video! I'm working with this simulator on my thesis and your videos are helping me out alot. One thing I wanted to address is that you are confused on why we would like to save the images of the sensors to disk. Well, I will use the simulator not for self-driving but for data acquisition (rgb, depth and segmentation data), so this method actually helps me out a bunch since then I will not have write my own code to read data and to then save it to disk. Just wanted to point that out! Once again thanks for the videos!
@Jmoss7
6 ай бұрын
Also working on a thesis project using carla, and this tutorial series is awesome so far!
@Watashiwapitadesu
6 ай бұрын
@@Jmoss7 Good luck! What is your topic?
@Jmoss7
6 ай бұрын
@@Watashiwapitadesu thank you! Just getting started, but the idea is simulating ethical dilemmas using Carla, and modeling them using the ODD (overview, design concepts, details) protocol. We’ll see how it pans out 😅
@shahadsabeeh3106
2 жыл бұрын
How can store steering angle, acceleration, and brake with each frame?
@shibz7866
3 жыл бұрын
Can you do a session on editing objects in Carla with UE4?
@molality
2 ай бұрын
Does this thing still work I guess I am pretty late homies
@goldenhok
5 жыл бұрын
Where is the part 2 of the text to speech neural net? I'm very excited to see how you would do a normal attention instead of a guided attention, i tested the code a little bit, but couldn't get any further a fully automated voice generator would be so useful, i'm tired of those robot voices in tutorials
@RohanOnBike
5 жыл бұрын
Awesome one sir. I am following ur channel for a long time now, infact ive. Learnt python from ur videos and website....also tried to recreate the gta v thing..but i did it on a small web based car racing game, and trained it enough to take turns and all. Can you please make a few videos on some competitive programming with python, like all the codechef, code forces, google code jam and other dynamic programming stuff. Im sure there are other many guyz who might feel this helpful, coz are really no videos. All comp. Programmers are mostly using c++.
@aravinddoss4127
5 жыл бұрын
Great work! I am also working with CARLA for my thesis. I had learned a lot from you and I will be glad to reciprocate. Expecting more videos from you!!!
@rezwanakabir8914
5 жыл бұрын
Hi aravind I have started learning Carla for course study. May I ask you about how to use it, what kind of data it needs and what kind of result it gives?
@vatsal2189
5 жыл бұрын
How did you know all the api calls? You learned them really quick
@Stinosko
5 жыл бұрын
Yeay part 2! :-D
@merth17
5 жыл бұрын
Stinosko how did he do that
@Stinosko
5 жыл бұрын
@@merth17 how did you what? 😉
@ADASDevelopment
Жыл бұрын
(carla-sim) C:\CARLA_0.9.14\PythonAPI\examples>python camera_test.py destroying actors done. Traceback (most recent call last): File "camera_test.py", line 36, in world = client.get_world() RuntimeError: time-out of 2000ms while waiting for the simulator, make sure the simulator is ready and connected to localhost:2000
@chadj1797
5 жыл бұрын
Yay! He is back :D
@sentdex
5 жыл бұрын
Where'd I go?
@chadj1797
5 жыл бұрын
@@sentdex Back as in with another video!
@SajanKumar-mx2jg
5 жыл бұрын
Was waiting for this only. Thanks for listening us!! Love you❤️
@Y0y0Jester
5 жыл бұрын
sentdex, how many different coffee mugs do you have in your house? Great video, as always btw!
@sentdex
5 жыл бұрын
Over 9000!!
@shivanisisodiya1438
2 жыл бұрын
Hello sentdex, I have a question. As I'm new to python, my thesis is dynamic risk assesment . Will you please tell me if that is the one you are talking about. That would be really helpful! Thanks.
@srikanthsrinivas8797
5 жыл бұрын
I think the save_to_disk method is to save images for training in a different environment later. For example using the images to train a CNN for steering. I think raw_data is what you need to directly pass raw_data to your model while actually testing your model. Maybe? Or maybe IDK what I'm talking about lol
@leedida2134
2 жыл бұрын
thanks for everything, this tutorials will help my project. thanks! you are really nice tutor for me.
@blanamaxima
4 жыл бұрын
most people should save the image, we train offline not online. Most likely they copy the framebuffer and that one is rgba.
@asdfasdf-vp9up
5 жыл бұрын
do you follow comma ai, they have some great self driving tech. you should look in to how they do self driving cars
@sentdex
5 жыл бұрын
I am aware of them. They collect data from real drivers who use their app. I don't totally feel like doing all that :P
@asdfasdf-vp9up
5 жыл бұрын
@@sentdex but collect data from npc maybe if that is possible
@fetullahatas3927
4 жыл бұрын
save to disk is useful method if you are trying to collect some dataset ,
@namantuli1357
4 жыл бұрын
dll load failed when using Carla 9.5 with python 3.6 Can someone help!
@anna-mariasourelli2753
4 жыл бұрын
Hi ! Great video thanks for uploading! I'm completely new to this stuff and I am trying to work out if I can use carla to generate user-defined speeds and trajectories during specific manoeuvres for self-driving cars.. Any recommendations? Thanks so much!!
@eastwoodsamuel4
3 жыл бұрын
Do you try out the codes and programs before recording or you learn on the fly during the video?? Love hearing your process of learning a new technology. Cheers. Great series btw.
@lasredchris
5 жыл бұрын
Sensor data from the car Reward function Interface w Carla w python Set a timeout
@gianlucavernia9444
5 жыл бұрын
Repost: Greetings sentdex I am following your machine learning series and I was wondering if Mean Shift is used or could be used to for stocks. What do you think?
@shaz-z506
5 жыл бұрын
Please create a video on how do learn and how do keep yourself update with the latest stuff.
@SleepyBoBos
4 жыл бұрын
Brilliant! How many thousands of people will this inspire to contribute in the area of AI. There's a big push for everyone to know how to 'code'. It should be 'code AI' and you sir are contributing to this. Thank you
@xiangjiaovv1416
5 жыл бұрын
hi, how to import carla? My environment is the same as the@sentdex!
@thegreatskywalker
5 жыл бұрын
Thank you!!! I'm so excited about this series :)
@threads25
4 жыл бұрын
destroying actors done. Traceback (most recent call last): File ".\ztest1.py", line 36, in world = client.get_world() RuntimeError: time-out of 2000ms while waiting for the simulator, make sure the simulator is ready and connected to localhost:2000 hey i'm getting this error what should I do now...
@ViniciusHSAraujo
4 жыл бұрын
I have changed from "localhost" to "127.0.0.1" and solved this problem!
@apoorvwatsky
5 жыл бұрын
32:00 when I execute the code, I only see a blank white window. The code is same, I've checked. Maybe some issuess with opencv? Or is it because of python version? I use python 3.7 but currently I'm using python 3.5 because that's what is supported by Carla.
@sentdex
5 жыл бұрын
Imma wager...code isnt the same Have you tried literally copying and pasting from the text tutorial linked in desc?
@apoorvwatsky
5 жыл бұрын
@@sentdex Not entirely. I've followed the code in this video. This is what I'm doing. ideone.com/5sN1yu Couldn't use f-string in set_attribute() method because of py3.5 and was having issues with format(), not a big deal. Where am I going wrong?
@sentdex
5 жыл бұрын
Copy and paste from the tutorial, use format or whatever in place of f strings. If that doesnt work, come to discord.gg/sentdex and well get you set
@apoorvwatsky
5 жыл бұрын
@@sentdex It's not working as expected. May I know what should I do in the channel? I don't use discord that much.
@ghl295
4 жыл бұрын
I had the same issue in Ubuntu16.04, with Python version 3.5 and Carla version 0.9.6. And I moved to Windows10, run the same code with Python version 3.7 and Carla version 0.9.5 (which is same as sentdex did), and everything went fine. So I think this issue might be because of the compatibility between ubuntu and opencv-python. In ubuntu, everything worked well except for visualizing the images in real time. I hope this could be helpful for you.
@divyanshusahu6413
4 жыл бұрын
I thought of doing this yesterday night whilst watching david silvers lectures. Woke up this morning and BOOM. DUDE UR MY SAVIOUR
@lasredchris
5 жыл бұрын
Import Model 3 Gta - hardcoded - game engine
@mineburgerl804
5 жыл бұрын
@sentdex how are you planning on getting the data?
@AUdominoes
5 жыл бұрын
I’m also working on something similar at A&M with a team this summer - good to see someone trying out RL on the latest CARLA Api. Out of curiosity, what GPU are you using/planning to use for training?
@drdavidwallis
5 жыл бұрын
What version of opencv are you using?
@peltolik5194
4 жыл бұрын
34:06 It's Gleb Zarin from Mother Russia ))
@ducnguyen4973
4 жыл бұрын
Great tutorial. Big thanks !!!
@vaibhavacharya9428
5 жыл бұрын
Django Channels much needed
@alexthunderrex
2 жыл бұрын
Your tutorials are so 😍
@Totial
5 жыл бұрын
Charles has a new teacher and her name is Carla
@deeksharajvanshi4433
3 жыл бұрын
Very nice video with proper breakdown of little pieces of information. Tried running the same, I see the camera window but the vehicle does not move. Would appreciate your help. Thanks
@MostafaSameer_2015
11 ай бұрын
I see the same situtation, did you solve the issue?
@katonk3745
Ай бұрын
@@MostafaSameer_2015 the car is moving if you check it in spectator mode.The camera is not synchronized with the car.I also have the same issue
@Jahaniam
5 жыл бұрын
Good Job! Btw, since image.raw_data is flattened you can use "np.fromiter" instead of "np.array" :) It is ~2x faster
@barsozkan63
3 жыл бұрын
Can you explain this line? In this tutorial we have "i = np.array(image.raw_data)" , how can update this line with np.fromiter. Thanks :)
@Napknbook
4 жыл бұрын
Amazing content, thank you!
@manirajan__
5 жыл бұрын
Thanks for this series, do you have plan to use ROS.
@sentdex
5 жыл бұрын
Not atm, no
@ariquciha58
4 жыл бұрын
This is amazing. This deserves more views.
@BlakeEdwards333
5 жыл бұрын
Was waiting for part 2! Thanks!!!
@NeoCodingGod
5 жыл бұрын
Can't get the notification bell to stick on your channel! Every time I set it & then find out it's unset!!!!
@theseamusexperience
5 жыл бұрын
Hi cool video. It is quite long though. I would suggest editing out some of the dead air time, it would improve watchability. Do you air initially as a live stream? I could totally see this format working
@gshan994
5 жыл бұрын
Swnrdex, How we can use the Reinforcement Learning model from carla in a real life situation such as using rpi for autonomous vehicles
@Shah_Khan
5 жыл бұрын
How do you keep yourself so much motivated to do these things? please answer this one.
@jeremiahblessing3891
5 жыл бұрын
Hello sentdex, I think it would be great if you implement Deep learning in to ROS [ Robot Operating System]. This will not only limit to the cars. You can automate anything[like robots, cars anything] and they behaviors can be trained in neural networks. ROS can be programmed in python. I think every one will benefit towards this[especially people who wish to work as robotics engineers] . Thank you.
@sentdex
5 жыл бұрын
I don't currently understand why I might use ROS over something like linux. Feel free to educate me :)
@jeremiahblessing3891
5 жыл бұрын
@@sentdex no no. ROS is a framework for robotics(it's not an operating system) . It runs inside Linux.. You've reviewed the Nvidia jetson in one of your video.. Actually robotics engineers use these Jetson along with ROS to control robots and now with the power of Jetson, the neural network can be introduced.. That's what I said
@jeremiahblessing3891
5 жыл бұрын
@@sentdex my point is you have a lot of machine learning related content. I think it would be great if you embed these with the less spoken ROS but widely used by professional robotic engineers.
@sentdex
5 жыл бұрын
It's not an operating system? Isn't the OS in ROS..."operating system?"
@jeremiahblessing3891
5 жыл бұрын
@@sentdex no, It was named that way. It's like a "brain" of the robot. That's why it's called as "operating" system.
@ImaginarySpace
5 жыл бұрын
Took like 5 seconds to find carla image raw_data and all other stuff explained by typing "image" to the documentation search... If you do programming you should first learn to use search and then everything else...
@sentdex
5 жыл бұрын
5 seconds? Wow, that's a pathetic time mate. I found it in
Пікірлер: 147