Closed LMJayasundara closed 4 years ago
/nodejs.org/api/cli.html#cli_unhanode train.js --data C:/Users/SHAN/Desktop/Tensor/Data --epochs 5 --model C:/Users/SHAN/Desktop/Tensor/model
Overriding the gradient for 'Max'
Overriding the gradient for 'OneHot'
Overriding the gradient for 'PadV2'
Overriding the gradient for 'SpaceToBatchND'
Overriding the gradient for 'SplitV'
2020-06-18 03:18:47.658620: I tensorflow/core/platform/cpu_feature_guard.cc:142] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2
Epoch 1 / 5
(node:1984) UnhandledPromiseRejectionWarning: Error: ENOENT: no such file or directory, open 'C:\Users\SHAN\Desktop\Tensor\ C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\left_2020_05_03_11_51_44_721.jpg'
(node:1984) UnhandledPromiseRejectionWarning: Unhandled promise rejection. This error originated either by throwing inside of an async function without a catch block, or by rejecting a promise which was not handled with .catch(). To terminate the node process on unhandled promise rejection, use the CLI flag --unhandled-rejections=strict
(see https://nodejs.org/api/cli.html#cli_unhandled_rejections_mode). (rejection id: 1)
(node:1984) [DEP0018] DeprecationWarning: Unhandled promise rejections are deprecated. In the future, promise rejections that are not handled will terminate the Node.js process
with a non-zero exit code.
(node:1984) UnhandledPromiseRejectionWarning: Error: ENOENT: no such file or directory, open 'C:\Users\SHAN\Desktop\Tensor\ C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\right_2020_05_03_11_51_44_721.jpg'
(node:1984) UnhandledPromiseRejectionWarning: Unhandled promise rejection. This error originated either by throwing inside of an async function without a catch block, or by rejecting a promise which was not handled with .catch(). To terminate the node process on unhandled promise rejection, use the CLI flag --unhandled-rejections=strict
(see https://nodejs.org/api/cli.html#cli_unhandled_rejections_mode). (rejection id: 2)
(node:1984) PromiseRejectionHandledWarning: Promise rejection was handled asynchronously (rejection id: 1)
(node:1984) UnhandledPromiseRejectionWarning: Error: ENOENT: no such file or directory, open 'C:\Users\SHAN\Desktop\Tensor\ C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\left_2020_05_03_11_51_44_721.jpg'
(node:1984) UnhandledPromiseRejectionWarning: Unhandled promise rejection. This error originated either by throwing inside of an async function without a catch block, or by rejecting a promise which was not handled with .catch(). To terminate the node process on unhandled promise rejection, use the CLI flag --unhandled-rejections=strict
(see https://nodejs.org/api/cli.html#cli_unhandled_rejections_mode). (rejection id: 4)
Could you please share a few first lines of driving_log.csv
generated by the simulator?
C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\center_2020_05_03_11_51_44_721.jpg, C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\left_2020_05_03_11_51_44_721.jpg, C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\right_2020_05_03_11_51_44_721.jpg, 0, 0, 0, 7.838477E-05 C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\center_2020_05_03_11_51_45_471.jpg, C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\left_2020_05_03_11_51_45_471.jpg, C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\right_2020_05_03_11_51_45_471.jpg, 0, 0, 0, 7.862586E-05 C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\center_2020_05_03_11_51_45_783.jpg, C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\left_2020_05_03_11_51_45_783.jpg, C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\right_2020_05_03_11_51_45_783.jpg, 0, 0, 0, 7.828791E-05 C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\center_2020_05_03_11_51_45_880.jpg, C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\left_2020_05_03_11_51_45_880.jpg, C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\right_2020_05_03_11_51_45_880.jpg, 0, 0, 0, 7.934251E-05 C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\center_2020_05_03_11_51_45_993.jpg, C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\left_2020_05_03_11_51_45_993.jpg, C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\right_2020_05_03_11_51_45_993.jpg, 0, 0, 0, 7.79378E-05 C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\center_2020_05_03_11_51_46_131.jpg, C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\left_2020_05_03_11_51_46_131.jpg, C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\right_2020_05_03_11_51_46_131.jpg, 0, 0, 0, 7.823821E-05 C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\center_2020_05_03_11_51_46_647.jpg, C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\left_2020_05_03_11_51_46_647.jpg, C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\right_2020_05_03_11_51_46_647.jpg, 0, 0, 0, 8.101902E-05 C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\center_2020_05_03_11_51_46_748.jpg, C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\left_2020_05_03_11_51_46_748.jpg, C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\right_2020_05_03_11_51_46_748.jpg, 0, 0, 0, 7.787524E-05
Could you please explain the environment you are using to run the script?
What is /nodejs.org/api/cli.html#cli_unhanode
?
nodejs v12.18.0 npm 6.14.4
do you have any solution
Now it should be fixed. Feel free to reopen otherwise. Thanks for reporting this issue.
Thank you..
Now it's working. can you add the IMG data path in the separate directory without using a simulator directory? can you modify your code in the above way???
Thank you for your great help.
On Thu, Jun 18, 2020 at 11:06 PM Alex Bakoushin notifications@github.com wrote:
Now it should be fixed. Feel free to reopen otherwise. Thanks for reporting this issue.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-646207289, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFAGNR5XLECERKRPHBD3RXJGBFANCNFSM4OA7SE5A .
Why do we need this? The IMG path is generated by the simulator and this is the exact path where the image is stored.
we can store the IMG folder in the same Code folder. It is easy to handle the code.
On Fri, Jun 19, 2020 at 12:41 PM Alex Bakoushin notifications@github.com wrote:
Why do we need this? The IMG path is generated by the simulator and this is the exact path where the image is stored.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-646477475, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFAEELETW7DY3HBOAPXDRXMFSPANCNFSM4OA7SE5A .
If you can remove the head path of the load png image that would be work. I try to do that. but ima fail. please can you help me?
I am using that
var path = require('path');
var filename = path.basename('/Users/Refsnes/demo_path.js'); console.log(filename); Run example » https://www.w3schools.com/nodejs/shownodejs_cmd.asp?filename=demo_met_path_basename
On Fri, Jun 19, 2020 at 1:06 PM Shan s.s.shancoin@gmail.com wrote:
we can store the IMG folder in the same Code folder. It is easy to handle the code.
On Fri, Jun 19, 2020 at 12:41 PM Alex Bakoushin notifications@github.com wrote:
Why do we need this? The IMG path is generated by the simulator and this is the exact path where the image is stored.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-646477475, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFAEELETW7DY3HBOAPXDRXMFSPANCNFSM4OA7SE5A .
If you want to construct a specific path, use path.join()
. Example:
const path = require('path');
const initialPath = 'C:\\Users\\SHAN\\Desktop\\simulator-windows-64\\Data\\IMG\\center_2020_05_03_11_51_46_131.jpg';
const filename = path.basename(initialPath);
const newPath = path.join(__dirname, 'IMG', filename);
console.log(newPath);
Thank you... how to add this part you train.js code ☺
On Sat, Jun 20, 2020 at 12:48 AM Alex Bakoushin notifications@github.com wrote:
If you want to construct a specific path, use path.join(). Example:
const path = require('path');const initialPath = 'C:\Users\SHAN\Desktop\simulator-windows-64\Data\IMG\center_2020_05_03_11_51_46_131.jpg';const filename = path.basename(initialPath)const newPath = path.join(__dirname, 'IMG', filename)console.log(newPath);
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-646825023, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFAAMCKRHVKD7GNJZ73DRXO2YVANCNFSM4OA7SE5A .
Why won't you do it yourself? 😄
That could be a good PR!
Nevertheless, here are the changes: https://github.com/bakoushin/self-driving-car-javascript/blob/relative-path/train.js#L34-L42
Just checkout branch named relative-path
.
What it does: instead of using absolute paths to the images stored in driving-log.csv
, it assumes now that images are stored in the IMG
directory within the data
directory we've supplied to the script.
So now we can move data directory anywhere we want, everything should work.
I hope this is what you are looking for.
yeaaaaaaaaaaaaaaaaaaaa... thank you sooooooooooooooooooooo much,...😍☺😍
On Sat, Jun 20, 2020 at 2:02 PM Alex Bakoushin notifications@github.com wrote:
Why won't you do it yourself? 😄 That could be a good PR!
Nevertheless, here are the changes: https://github.com/bakoushin/self-driving-car-javascript/blob/relative-path/train.js#L34-L42 Just checkout branch named relative-path.
What it does: instead of using absolute paths to the images stored in driving-log.csv, it assumes now that images are stored in the IMG directory within the data directory we've supplied to the script. So now we can move data directory anywhere we want, everything should work.
I hope this is what you are looking for.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-646963170, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFAEW4SHWAZ5DC6XRAXLRXRXZZANCNFSM4OA7SE5A .
Can I connect Carla simulator in a similar way?
On Sat, Jun 20, 2020 at 3:19 PM Shan s.s.shancoin@gmail.com wrote:
yeaaaaaaaaaaaaaaaaaaaa... thank you sooooooooooooooooooooo much,...😍☺😍
On Sat, Jun 20, 2020 at 2:02 PM Alex Bakoushin notifications@github.com wrote:
Why won't you do it yourself? 😄 That could be a good PR!
Nevertheless, here are the changes: https://github.com/bakoushin/self-driving-car-javascript/blob/relative-path/train.js#L34-L42 Just checkout branch named relative-path.
What it does: instead of using absolute paths to the images stored in driving-log.csv, it assumes now that images are stored in the IMG directory within the data directory we've supplied to the script. So now we can move data directory anywhere we want, everything should work.
I hope this is what you are looking for.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-646963170, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFAEW4SHWAZ5DC6XRAXLRXRXZZANCNFSM4OA7SE5A .
I connected carala with js. But I cannot control it
var net = require('net');
var client = new net.Socket(); client.connect(2000, 'localhost', function() { console.log('Connected'); client.setTimeout(10.0); });
On Mon, Jun 22, 2020 at 8:59 AM Shan s.s.shancoin@gmail.com wrote:
Can I connect Carla simulator in a similar way?
On Sat, Jun 20, 2020 at 3:19 PM Shan s.s.shancoin@gmail.com wrote:
yeaaaaaaaaaaaaaaaaaaaa... thank you sooooooooooooooooooooo much,...😍☺😍
On Sat, Jun 20, 2020 at 2:02 PM Alex Bakoushin notifications@github.com wrote:
Why won't you do it yourself? 😄 That could be a good PR!
Nevertheless, here are the changes: https://github.com/bakoushin/self-driving-car-javascript/blob/relative-path/train.js#L34-L42 Just checkout branch named relative-path.
What it does: instead of using absolute paths to the images stored in driving-log.csv, it assumes now that images are stored in the IMG directory within the data directory we've supplied to the script. So now we can move data directory anywhere we want, everything should work.
I hope this is what you are looking for.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-646963170, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFAEW4SHWAZ5DC6XRAXLRXRXZZANCNFSM4OA7SE5A .
I don't know. Could you provide a link to Carla simulator?
Yes of course. https://carla.org/ This is Carla simulator web page. It is a python based simulator environment.
I just download the Carla simulator then opened exe file and run js code ant it was connected. 😍😍 As I mention earlier.
It has some python code for simulated a car in that environment. So have a very straightforward youtube video that describes how to simulate a car in that environment in python.
simulate video - https://www.youtube.com/watch?v=2hM44nr7Wms&list=PLQVvvaa0QuDeI12McNQdnTlWz9XlCa0uo&index=3&t=0s playlist - https://www.youtube.com/playlist?list=PLQVvvaa0QuDeI12McNQdnTlWz9XlCa0uo
also, this is very useful to get an idea - https://carla.org/2018/07/30/release-0.9.0/
If we can develop this environment using Js it will very useful for a simulate big project with TensorFlow Js.💪💪
On Mon, Jun 22, 2020 at 12:03 PM Alex Bakoushin notifications@github.com wrote:
I don't know. Could you provide a link to Carla simulator?
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-647314956, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFABMB64266QOZ5US43LRX33KZANCNFSM4OA7SE5A .
Did you get an idea?☺
On Mon, Jun 22, 2020 at 2:02 PM Shan s.s.shancoin@gmail.com wrote:
Yes of course. https://carla.org/ This is Carla simulator web page. It is a python based simulator environment.
I just download the Carla simulator then opened exe file and run js code ant it was connected. 😍😍 As I mention earlier.
It has some python code for simulated a car in that environment. So have a very straightforward youtube video that describes how to simulate a car in that environment in python.
simulate video - https://www.youtube.com/watch?v=2hM44nr7Wms&list=PLQVvvaa0QuDeI12McNQdnTlWz9XlCa0uo&index=3&t=0s playlist - https://www.youtube.com/playlist?list=PLQVvvaa0QuDeI12McNQdnTlWz9XlCa0uo
also, this is very useful to get an idea - https://carla.org/2018/07/30/release-0.9.0/
If we can develop this environment using Js it will very useful for a simulate big project with TensorFlow Js.💪💪
On Mon, Jun 22, 2020 at 12:03 PM Alex Bakoushin notifications@github.com wrote:
I don't know. Could you provide a link to Carla simulator?
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-647314956, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFABMB64266QOZ5US43LRX33KZANCNFSM4OA7SE5A .
I also tried to implement a speed based steering angle correction with udacity simulator
On Mon, Jun 22, 2020 at 8:57 PM Shan s.s.shancoin@gmail.com wrote:
Did you get an idea?☺
On Mon, Jun 22, 2020 at 2:02 PM Shan s.s.shancoin@gmail.com wrote:
Yes of course. https://carla.org/ This is Carla simulator web page. It is a python based simulator environment.
I just download the Carla simulator then opened exe file and run js code ant it was connected. 😍😍 As I mention earlier.
It has some python code for simulated a car in that environment. So have a very straightforward youtube video that describes how to simulate a car in that environment in python.
simulate video - https://www.youtube.com/watch?v=2hM44nr7Wms&list=PLQVvvaa0QuDeI12McNQdnTlWz9XlCa0uo&index=3&t=0s playlist - https://www.youtube.com/playlist?list=PLQVvvaa0QuDeI12McNQdnTlWz9XlCa0uo
also, this is very useful to get an idea - https://carla.org/2018/07/30/release-0.9.0/
If we can develop this environment using Js it will very useful for a simulate big project with TensorFlow Js.💪💪
On Mon, Jun 22, 2020 at 12:03 PM Alex Bakoushin notifications@github.com wrote:
I don't know. Could you provide a link to Carla simulator?
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-647314956, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFABMB64266QOZ5US43LRX33KZANCNFSM4OA7SE5A .
can you get any solution? I try but I can not get solution🤕🤕
On Mon, Jun 22, 2020 at 9:14 PM Shan s.s.shancoin@gmail.com wrote:
I also tried to implement a speed based steering angle correction with udacity simulator
On Mon, Jun 22, 2020 at 8:57 PM Shan s.s.shancoin@gmail.com wrote:
Did you get an idea?☺
On Mon, Jun 22, 2020 at 2:02 PM Shan s.s.shancoin@gmail.com wrote:
Yes of course. https://carla.org/ This is Carla simulator web page. It is a python based simulator environment.
I just download the Carla simulator then opened exe file and run js code ant it was connected. 😍😍 As I mention earlier.
It has some python code for simulated a car in that environment. So have a very straightforward youtube video that describes how to simulate a car in that environment in python.
simulate video - https://www.youtube.com/watch?v=2hM44nr7Wms&list=PLQVvvaa0QuDeI12McNQdnTlWz9XlCa0uo&index=3&t=0s playlist - https://www.youtube.com/playlist?list=PLQVvvaa0QuDeI12McNQdnTlWz9XlCa0uo
also, this is very useful to get an idea - https://carla.org/2018/07/30/release-0.9.0/
If we can develop this environment using Js it will very useful for a simulate big project with TensorFlow Js.💪💪
On Mon, Jun 22, 2020 at 12:03 PM Alex Bakoushin < notifications@github.com> wrote:
I don't know. Could you provide a link to Carla simulator?
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-647314956, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFABMB64266QOZ5US43LRX33KZANCNFSM4OA7SE5A .
Thanks for providing the information about Carla.
It is way more advanced than the Udacity simulator in terms of the complexity of simulated word and control APIs as well. In short, this project would not work with Carla unless it is completely re-implemented for it.
Since the Carla environment is way rich that Udacity one, more sophisticated models may be required to correctly drive a car. And when it comes to experimenting with different models, the Python infrastructure is the natural choice given the abundance of established tools and libraries, TensorFlow included.
okay, Thank you for your advice☺💪
On Wed, Jun 24, 2020 at 12:31 PM Alex Bakoushin notifications@github.com wrote:
Thanks for providing the information about Carla.
It is way more advanced than the Udacity simulator in terms of the complexity of simulated word and control APIs as well. In short, this project would not work with Carla unless it is completely re-implemented for it.
Since the Carla environment is way rich that Udacity one, more sophisticated models may be required to correctly drive a car. And when it comes to experimenting with different models, the Python infrastructure is the natural choice given the abundance of established tools and libraries, TensorFlow included.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-648633530, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFAHYN5JLAIH2BHNEY7DRYGQDJANCNFSM4OA7SE5A .
I am trying to get an image from the camera and create a CSV file like that Udasity simulator. Also, get steering and throttle(But not now) This way I can train a real wold robot car.😁
This way create a CSV file But it can not include any data it always empty. Please show me the error I did
const cv = require('/opncvnode/node_modules/opencv4nodejs'); const fs = require('fs'); const csv = require('fast-csv'); const ws = fs.createWriteStream('test.csv');
var sec = 0; var frameRate = 0.5; var count = 1;
var cap = new cv.VideoCapture(0);
function getFrame(sec){ cap.set(cv.CAP_PROP_POS_MSEC, sec*1000); var frame = cap.read();
cv.imwrite("image/center_cam" + count + ".jpg", frame);
cv.imshow('',frame);
cv.waitKey(1);
///////////////////////////////////////////////////////////////////////////////////////
var image = "center_cam"+count+".jpg";
var steering = 0;
var throttle = 0;
console.log(image, steering, throttle);
var data = [{image, steering, throttle}];
csv.write(data, { headers: true })
.pipe(ws);
///////////////////////////////////////////////////////////////////////////////////////
return frame;
};
var success = getFrame(sec);
while(success = true){ var count = count + 1; var sec = sec + frameRate; var sec = Math.round(sec); var success = getFrame(sec); };
This way create a CSV file also write data
for (i=0; i<10; i++){
var num1 = Math.random(); var num2 = Math.random(); var add = num1 + num2;
var data = [{num1, num2, add}];
csv.write(data, { headers: true }) .pipe(ws);
}
On Wed, Jun 24, 2020 at 1:09 PM Shan s.s.shancoin@gmail.com wrote:
okay, Thank you for your advice☺💪
On Wed, Jun 24, 2020 at 12:31 PM Alex Bakoushin notifications@github.com wrote:
Thanks for providing the information about Carla.
It is way more advanced than the Udacity simulator in terms of the complexity of simulated word and control APIs as well. In short, this project would not work with Carla unless it is completely re-implemented for it.
Since the Carla environment is way rich that Udacity one, more sophisticated models may be required to correctly drive a car. And when it comes to experimenting with different models, the Python infrastructure is the natural choice given the abundance of established tools and libraries, TensorFlow included.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-648633530, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFAHYN5JLAIH2BHNEY7DRYGQDJANCNFSM4OA7SE5A .
But it also gets wrong type data format
num1,num2,add 0.7260764056613853,0.2969767742939857,1.023053179955371num1,num2,add 0.9313625225535354,0.7464783328796931,1.6778408554332285num1,num2,add 0.9055004970764169,0.7301077704172396,1.6356082674936565num1,num2,add 0.18310182726664648,0.2515544687958069,0.4346562960624534num1,num2,add 0.9159285448265637,0.5768440038442426,1.4927725486708063num1,num2,add 0.7040326362130156,0.7418406729086742,1.4458733091216898num1,num2,add 0.7971635785667379,0.5465111704901595,1.3436747490568974num1,num2,add 0.8682385206885539,0.03586618439733935,0.9041047050858932num1,num2,add 0.010650019081321904,0.561125578450826,0.5717755975321479num1,num2,add 0.7253343112313102,0.642651626168383,1.3679859373996932
On Mon, Jun 29, 2020 at 11:11 AM Shan s.s.shancoin@gmail.com wrote:
I am trying to get an image from the camera and create a CSV file like that Udasity simulator. Also, get steering and throttle(But not now) This way I can train a real wold robot car.😁
This way create a CSV file But it can not include any data it always empty. Please show me the error I did
const cv = require('/opncvnode/node_modules/opencv4nodejs'); const fs = require('fs'); const csv = require('fast-csv'); const ws = fs.createWriteStream('test.csv');
var sec = 0; var frameRate = 0.5; var count = 1;
var cap = new cv.VideoCapture(0);
function getFrame(sec){ cap.set(cv.CAP_PROP_POS_MSEC, sec*1000); var frame = cap.read();
cv.imwrite("image/center_cam" + count + ".jpg", frame); cv.imshow('',frame); cv.waitKey(1);
///////////////////////////////////////////////////////////////////////////////////////
var image = "center_cam"+count+".jpg"; var steering = 0; var throttle = 0; console.log(image, steering, throttle); var data = [{image, steering, throttle}]; csv.write(data, { headers: true }) .pipe(ws);
///////////////////////////////////////////////////////////////////////////////////////
return frame;
};
var success = getFrame(sec);
while(success = true){ var count = count + 1; var sec = sec + frameRate; var sec = Math.round(sec); var success = getFrame(sec); };
This way create a CSV file also write data
for (i=0; i<10; i++){
var num1 = Math.random(); var num2 = Math.random(); var add = num1 + num2;
var data = [{num1, num2, add}];
csv.write(data, { headers: true }) .pipe(ws);
}
On Wed, Jun 24, 2020 at 1:09 PM Shan s.s.shancoin@gmail.com wrote:
okay, Thank you for your advice☺💪
On Wed, Jun 24, 2020 at 12:31 PM Alex Bakoushin notifications@github.com wrote:
Thanks for providing the information about Carla.
It is way more advanced than the Udacity simulator in terms of the complexity of simulated word and control APIs as well. In short, this project would not work with Carla unless it is completely re-implemented for it.
Since the Carla environment is way rich that Udacity one, more sophisticated models may be required to correctly drive a car. And when it comes to experimenting with different models, the Python infrastructure is the natural choice given the abundance of established tools and libraries, TensorFlow included.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-648633530, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFAHYN5JLAIH2BHNEY7DRYGQDJANCNFSM4OA7SE5A .
I also ask StackOverflow and GitHub but I can not get solution for that Please please send me what I did wrong.
Thank you
😃
On Mon, Jun 29, 2020 at 11:13 AM Shan s.s.shancoin@gmail.com wrote:
But it also gets wrong type data format
num1,num2,add 0.7260764056613853,0.2969767742939857,1.023053179955371num1,num2,add 0.9313625225535354,0.7464783328796931,1.6778408554332285num1,num2,add 0.9055004970764169,0.7301077704172396,1.6356082674936565num1,num2,add 0.18310182726664648,0.2515544687958069,0.4346562960624534num1,num2,add 0.9159285448265637,0.5768440038442426,1.4927725486708063num1,num2,add 0.7040326362130156,0.7418406729086742,1.4458733091216898num1,num2,add 0.7971635785667379,0.5465111704901595,1.3436747490568974num1,num2,add 0.8682385206885539,0.03586618439733935,0.9041047050858932num1,num2,add 0.010650019081321904,0.561125578450826,0.5717755975321479num1,num2,add 0.7253343112313102,0.642651626168383,1.3679859373996932
On Mon, Jun 29, 2020 at 11:11 AM Shan s.s.shancoin@gmail.com wrote:
I am trying to get an image from the camera and create a CSV file like that Udasity simulator. Also, get steering and throttle(But not now) This way I can train a real wold robot car.😁
This way create a CSV file But it can not include any data it always empty. Please show me the error I did
const cv = require('/opncvnode/node_modules/opencv4nodejs'); const fs = require('fs'); const csv = require('fast-csv'); const ws = fs.createWriteStream('test.csv');
var sec = 0; var frameRate = 0.5; var count = 1;
var cap = new cv.VideoCapture(0);
function getFrame(sec){ cap.set(cv.CAP_PROP_POS_MSEC, sec*1000); var frame = cap.read();
cv.imwrite("image/center_cam" + count + ".jpg", frame); cv.imshow('',frame); cv.waitKey(1);
///////////////////////////////////////////////////////////////////////////////////////
var image = "center_cam"+count+".jpg"; var steering = 0; var throttle = 0; console.log(image, steering, throttle); var data = [{image, steering, throttle}]; csv.write(data, { headers: true }) .pipe(ws);
///////////////////////////////////////////////////////////////////////////////////////
return frame;
};
var success = getFrame(sec);
while(success = true){ var count = count + 1; var sec = sec + frameRate; var sec = Math.round(sec); var success = getFrame(sec); };
This way create a CSV file also write data
for (i=0; i<10; i++){
var num1 = Math.random(); var num2 = Math.random(); var add = num1 + num2;
var data = [{num1, num2, add}];
csv.write(data, { headers: true }) .pipe(ws);
}
On Wed, Jun 24, 2020 at 1:09 PM Shan s.s.shancoin@gmail.com wrote:
okay, Thank you for your advice☺💪
On Wed, Jun 24, 2020 at 12:31 PM Alex Bakoushin < notifications@github.com> wrote:
Thanks for providing the information about Carla.
It is way more advanced than the Udacity simulator in terms of the complexity of simulated word and control APIs as well. In short, this project would not work with Carla unless it is completely re-implemented for it.
Since the Carla environment is way rich that Udacity one, more sophisticated models may be required to correctly drive a car. And when it comes to experimenting with different models, the Python infrastructure is the natural choice given the abundance of established tools and libraries, TensorFlow included.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-648633530, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFAHYN5JLAIH2BHNEY7DRYGQDJANCNFSM4OA7SE5A .
I believe since you are wrtiting to a Writable Stream you want to use writeToStream
const csv = require('fast-csv'); const csvStream = csv.format();
const fs = require('fs'); const ws = fs.createWriteStream('test.csv');
csvStream.pipe(ws).on('end', () => process.exit());
for(i = 0; i<10; i++){ var image = 'center'+i+'.jpg'; var steering = 0; csvStream.write({ image, steering }); }
csvStream.end();
I write this way it works!!!
But I can not include this to the main code it shows empty CSV file
On Mon, Jun 29, 2020 at 10:10 PM Alex Bakoushin notifications@github.com wrote:
I believe since you are wrtiting to a Writable Stream https://nodejs.org/api/fs.html#fs_fs_createwritestream_path_options you want to use writeToStream https://c2fo.io/fast-csv/docs/formatting/methods#writetostream
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-651234074, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFAAX4ZCDQ6XUWC64HZTRZC7ZDANCNFSM4OA7SE5A .
my main code
const cv = require('/opncvnode/node_modules/opencv4nodejs');
const csv = require('fast-csv'); const csvStream = csv.format();
const fs = require('fs'); const ws = fs.createWriteStream('test.csv');
csvStream.pipe(ws).on('end', () => process.exit());
var sec = 0; var frameRate = 0.5; var count = 1;
var cap = new cv.VideoCapture('video.mp4');
function getFrame(sec){
cap.set(cv.CAP_PROP_POS_MSEC, sec*1000);
var frame = cap.read();
var img_dir = "image/";
var img_name = "center_cam"+ count +".jpg";
var image = img_dir + img_name;
cv.imwrite(image, frame);
///////////////////////////////////////////////////////////////////
var steering = 0;
csvStream.write({ img_name, steering });
///////////////////////////////////////////////////////////////////
cv.imshow('', frame);
cv.waitKey(1);
return frame;
};
var success = getFrame(sec);
while(success = true){ var count = count + 1; var sec = sec + frameRate; var sec = Math.round(sec); var success = getFrame(sec); };
csvStream.end();
On Mon, Jun 29, 2020 at 10:36 PM Shan s.s.shancoin@gmail.com wrote:
const csv = require('fast-csv'); const csvStream = csv.format();
const fs = require('fs'); const ws = fs.createWriteStream('test.csv');
csvStream.pipe(ws).on('end', () => process.exit());
for(i = 0; i<10; i++){ var image = 'center'+i+'.jpg'; var steering = 0; csvStream.write({ image, steering }); }
csvStream.end();
I write this way it works!!!
But I can not include this to the main code it shows empty CSV file
On Mon, Jun 29, 2020 at 10:10 PM Alex Bakoushin notifications@github.com wrote:
I believe since you are wrtiting to a Writable Stream https://nodejs.org/api/fs.html#fs_fs_createwritestream_path_options you want to use writeToStream https://c2fo.io/fast-csv/docs/formatting/methods#writetostream
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-651234074, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFAAX4ZCDQ6XUWC64HZTRZC7ZDANCNFSM4OA7SE5A .
This could happen because your code is failing with some error somewhere. I suggest checking console output and getting rid of these errors first.
Some ideas:
Probably you want to check if frame.empty
before proceed.
Also, isn't that condition while(success = true)
a typo?
okay thank you so much I will check💪😍
On Tue, Jun 30, 2020 at 7:40 PM Alex Bakoushin notifications@github.com wrote:
This could happen because your code is failing with some error somewhere. I suggest checking console output and getting rid of these errors first.
Some ideas: Probably you want to check if frame.empty before proceed. Also, isn't that condition while(success = true) a typo?
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-651816543, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFADT5CNZ5EJAFA2HERLRZHW6PANCNFSM4OA7SE5A .
Finally I did it..💪😃😍💪😍
On Tue, Jun 30, 2020 at 8:41 PM Shan s.s.shancoin@gmail.com wrote:
okay thank you so much I will check💪😍
On Tue, Jun 30, 2020 at 7:40 PM Alex Bakoushin notifications@github.com wrote:
This could happen because your code is failing with some error somewhere. I suggest checking console output and getting rid of these errors first.
Some ideas: Probably you want to check if frame.empty before proceed. Also, isn't that condition while(success = true) a typo?
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-651816543, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFADT5CNZ5EJAFA2HERLRZHW6PANCNFSM4OA7SE5A .
Can I do image augmentation with Javascript/NodeJs like zooming change brightness and panning images
On Wed, Jul 1, 2020 at 2:36 AM Shan s.s.shancoin@gmail.com wrote:
Finally I did it..💪😃😍💪😍
On Tue, Jun 30, 2020 at 8:41 PM Shan s.s.shancoin@gmail.com wrote:
okay thank you so much I will check💪😍
On Tue, Jun 30, 2020 at 7:40 PM Alex Bakoushin notifications@github.com wrote:
This could happen because your code is failing with some error somewhere. I suggest checking console output and getting rid of these errors first.
Some ideas: Probably you want to check if frame.empty before proceed. Also, isn't that condition while(success = true) a typo?
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-651816543, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFADT5CNZ5EJAFA2HERLRZHW6PANCNFSM4OA7SE5A .
Totally. If you are using OpenCV just apply the same techniques people use in Python. There are also interesting Node.js packages for image processing: https://github.com/lovell/sharp https://www.npmjs.com/package/jimp
okay, thank you very much. I am using opencv4nodejs😍
I have another question😁
Can I generate the preprocessed images (like zooming flipping images) to feed and train the algorithm without saving the disk.
Act like as an image augment generator
On Wed, Jul 1, 2020 at 2:45 PM Alex Bakoushin notifications@github.com wrote:
Totally. If you are using OpenCV just apply the same techniques people use in Python. There are also interesting Node.js packages for image processing: https://github.com/lovell/sharp https://www.npmjs.com/package/jimp
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-652298804, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFAFNKQN4DLHFCQILVLLRZL5BVANCNFSM4OA7SE5A .
Yes. Just preprocess the image before returning it from generator 😄 Note that preprocessing will significantly slower the training.
Yep. I agree with you. But I am trying to train the algorithm using only center camera images. So I need an image generator to get better output from the algorithm even it slower training... I know you already know it 😂😆 I had a code with image preprocessing.generator. This is my code for image generator.👇 I think it better to include inside generator function. But I can not find to way include this in your code. Please help me😃
const cv = require('/opncvnode/node_modules/opencv4nodejs');
function zoom(image){ var col = image.cols; var row = image.rows;
var center = new cv.Point((col-1)/2.0, (row-1)/2.0);
var mat = new cv.getRotationMatrix2D(center, 0, 1.5);
var zoom_img = image.warpAffine(mat, new cv.Size(col, row));
return zoom_img;
}
function brightness(image){ var bright_img = image.mul(0.6); return bright_img; }
function pan(image){ var top = Math.floor(Math.random() 20) + 1; var bottom = Math.floor(Math.random() 20) + 1; var left = Math.floor(Math.random() 20) + 1; var right = Math.floor(Math.random() 20) + 1;
const color = new cv.Vec(0, 0, 0);
const pan_image = image.copyMakeBorder(top, bottom, left, right, cv.
BORDER_CONSTANT, color);
return pan_image;
}
function flip(image){ var flip_img = image.flip(1); return flip_img; }
function imageAugmentation(image){ var image = cv.imread('center_cam.png');
if (Math.random() * 1 > 0.5) var image = zoom(image);
if (Math.random() * 1 > 0.5) var image = brightness(image);
if (Math.random() * 1 > 0.5) var image = pan(image);
if (Math.random() * 1 > 0.5) var image = flip(image);
return image;
}
var image = imageAugmentation(image);
cv.imshowWait('', image);
On Fri, Jul 3, 2020 at 1:46 PM Alex Bakoushin notifications@github.com wrote:
Yes. Just preprocess the image before returning it from generator 😄 Note that preprocessing will significantly slower the training.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-653417402, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFACWA77EWF6OL3FBJLLRZWHUBANCNFSM4OA7SE5A .
"Yes. Just preprocess the image before returning it from generator 😄 " Thank you for your advise 👆 I did it thank you very much...
But I am getting preprocess image as a variable... Then how it converts to an array buffer
var image = imageAugmentation(image); // output preprocess image as Mat
// console.log(image);
var processImageBuffer = fs.promises.read(image); <= This not work I am tring diferant way but not work
yield [await centerImageBuffer, Number(steering)]; yield [await processImageBuffer, Number(steering)];
Please please give me an advice.. Thank you
On Fri, Jul 3, 2020 at 2:37 PM Shan s.s.shancoin@gmail.com wrote:
Yep. I agree with you. But I am trying to train the algorithm using only center camera images. So I need an image generator to get better output from the algorithm even it slower training... I know you already know it 😂😆 I had a code with image preprocessing.generator. This is my code for image generator.👇 I think it better to include inside generator function. But I can not find to way include this in your code. Please help me😃
const cv = require('/opncvnode/node_modules/opencv4nodejs');
function zoom(image){ var col = image.cols; var row = image.rows;
var center = new cv.Point((col-1)/2.0, (row-1)/2.0); var mat = new cv.getRotationMatrix2D(center, 0, 1.5); var zoom_img = image.warpAffine(mat, new cv.Size(col, row)); return zoom_img;
}
function brightness(image){ var bright_img = image.mul(0.6); return bright_img; }
function pan(image){ var top = Math.floor(Math.random() 20) + 1; var bottom = Math.floor(Math.random() 20) + 1; var left = Math.floor(Math.random() 20) + 1; var right = Math.floor(Math.random() 20) + 1;
const color = new cv.Vec(0, 0, 0); const pan_image = image.copyMakeBorder(top, bottom, left, right, cv.
BORDER_CONSTANT, color);
return pan_image;
}
function flip(image){ var flip_img = image.flip(1); return flip_img; }
function imageAugmentation(image){ var image = cv.imread('center_cam.png');
if (Math.random() * 1 > 0.5) var image = zoom(image); if (Math.random() * 1 > 0.5) var image = brightness(image); if (Math.random() * 1 > 0.5) var image = pan(image); if (Math.random() * 1 > 0.5) var image = flip(image); return image;
}
var image = imageAugmentation(image);
cv.imshowWait('', image);
On Fri, Jul 3, 2020 at 1:46 PM Alex Bakoushin notifications@github.com wrote:
Yes. Just preprocess the image before returning it from generator 😄 Note that preprocessing will significantly slower the training.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-653417402, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFACWA77EWF6OL3FBJLLRZWHUBANCNFSM4OA7SE5A .
Also, I did this way but it get an error
var image = imageAugmentation(image);
// console.log(image);
var processImageBuffer = new ArrayBuffer(image);
yield [await centerImageBuffer, Number(steering)]; yield [await processImageBuffer, Number(steering)];
Error
(node:50600) UnhandledPromiseRejectionWarning: Io::Imdecode - expected arg
0 to be a Buffer of Uint8 Values
(node:50600) UnhandledPromiseRejectionWarning: Unhandled promise rejection.
This error originated either by throwing inside of an async
function without a catch block, or by rejecting a promise which was not
handled with .catch(). To terminate the node process on unhandled
promise rejection, use the CLI flag --unhandled-rejections=strict
(see
https://nodejs.org/api/cli.html#cli_unhandled_rejections_mode).
(rejection id: 1)
(node:50600) [DEP0018] DeprecationWarning: Unhandled promise rejections are
deprecated. In the future, promise rejections that are not
handled will terminate the Node.js process with a non-zero exit code.
On Sat, Jul 4, 2020 at 10:54 PM Shan s.s.shancoin@gmail.com wrote:
"Yes. Just preprocess the image before returning it from generator 😄 " Thank you for your advise 👆 I did it thank you very much...
But I am getting preprocess image as a variable... Then how it converts to an array buffer
var image = imageAugmentation(image); // output preprocess image as Mat
// console.log(image);
var processImageBuffer = fs.promises.read(image); <= This not work I am tring diferant way but not work
yield [await centerImageBuffer, Number(steering)]; yield [await processImageBuffer, Number(steering)];
Please please give me an advice.. Thank you
On Fri, Jul 3, 2020 at 2:37 PM Shan s.s.shancoin@gmail.com wrote:
Yep. I agree with you. But I am trying to train the algorithm using only center camera images. So I need an image generator to get better output from the algorithm even it slower training... I know you already know it 😂😆 I had a code with image preprocessing.generator. This is my code for image generator.👇 I think it better to include inside generator function. But I can not find to way include this in your code. Please help me😃
const cv = require('/opncvnode/node_modules/opencv4nodejs');
function zoom(image){ var col = image.cols; var row = image.rows;
var center = new cv.Point((col-1)/2.0, (row-1)/2.0); var mat = new cv.getRotationMatrix2D(center, 0, 1.5); var zoom_img = image.warpAffine(mat, new cv.Size(col, row)); return zoom_img;
}
function brightness(image){ var bright_img = image.mul(0.6); return bright_img; }
function pan(image){ var top = Math.floor(Math.random() 20) + 1; var bottom = Math.floor(Math.random() 20) + 1; var left = Math.floor(Math.random() 20) + 1; var right = Math.floor(Math.random() 20) + 1;
const color = new cv.Vec(0, 0, 0); const pan_image = image.copyMakeBorder(top, bottom, left, right, cv.
BORDER_CONSTANT, color);
return pan_image;
}
function flip(image){ var flip_img = image.flip(1); return flip_img; }
function imageAugmentation(image){ var image = cv.imread('center_cam.png');
if (Math.random() * 1 > 0.5) var image = zoom(image); if (Math.random() * 1 > 0.5) var image = brightness(image); if (Math.random() * 1 > 0.5) var image = pan(image); if (Math.random() * 1 > 0.5) var image = flip(image); return image;
}
var image = imageAugmentation(image);
cv.imshowWait('', image);
On Fri, Jul 3, 2020 at 1:46 PM Alex Bakoushin notifications@github.com wrote:
Yes. Just preprocess the image before returning it from generator 😄 Note that preprocessing will significantly slower the training.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-653417402, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFACWA77EWF6OL3FBJLLRZWHUBANCNFSM4OA7SE5A .
Check what exact types each part of your program is expecting. For example, if some function expects a Buffer, do not feed it with Promise. Make sure it's a Buffer. The same goes for image formats. For instance, my training code expects to get a JPEG image. So you may want either return a JPEG after augmentation or change the code to consume a raw Buffer. In all that concerns input types, the documentation is your friend. Also, make sure that augmentation yields images with the same dimensions.
okay... thank you
On Mon, Jul 6, 2020 at 1:34 PM Alex Bakoushin notifications@github.com wrote:
Check what exact types each part of your program is expecting. For example, if some function expects a Buffer, do not feed it with Promise. Make sure it's a Buffer. The same goes for image formats. For instance, my training code expects to get a JPEG image. So you may want either return a JPEG after augmentation or change the code to consume a raw Buffer. In all that concerns input types, the documentation is your friend. Also, make sure that augmentation yields images with the same dimensions.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-654081943, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFAH525H2Y4MIRRU4ZFLR2GAQHANCNFSM4OA7SE5A .
Hi, I need help from you. I developed a semantic segmentation model with Tensorflow Js But it not reduce acc in training. Can you help me to fix it😍😍
On Mon, Jul 6, 2020 at 1:55 PM Shan s.s.shancoin@gmail.com wrote:
okay... thank you
On Mon, Jul 6, 2020 at 1:34 PM Alex Bakoushin notifications@github.com wrote:
Check what exact types each part of your program is expecting. For example, if some function expects a Buffer, do not feed it with Promise. Make sure it's a Buffer. The same goes for image formats. For instance, my training code expects to get a JPEG image. So you may want either return a JPEG after augmentation or change the code to consume a raw Buffer. In all that concerns input types, the documentation is your friend. Also, make sure that augmentation yields images with the same dimensions.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-654081943, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFAH525H2Y4MIRRU4ZFLR2GAQHANCNFSM4OA7SE5A .
I think it will help for Javascript self-driving car projects💟
On Tue, Aug 25, 2020 at 2:46 AM Shan s.s.shancoin@gmail.com wrote:
Hi, I need help from you. I developed a semantic segmentation model with Tensorflow Js But it not reduce acc in training. Can you help me to fix it😍😍
On Mon, Jul 6, 2020 at 1:55 PM Shan s.s.shancoin@gmail.com wrote:
okay... thank you
On Mon, Jul 6, 2020 at 1:34 PM Alex Bakoushin notifications@github.com wrote:
Check what exact types each part of your program is expecting. For example, if some function expects a Buffer, do not feed it with Promise. Make sure it's a Buffer. The same goes for image formats. For instance, my training code expects to get a JPEG image. So you may want either return a JPEG after augmentation or change the code to consume a raw Buffer. In all that concerns input types, the documentation is your friend. Also, make sure that augmentation yields images with the same dimensions.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bakoushin/self-driving-car-javascript/issues/1#issuecomment-654081943, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLQFAH525H2Y4MIRRU4ZFLR2GAQHANCNFSM4OA7SE5A .
I am trying to loading IMG file for the train.js It shows cannot identify the directory please let me know how to load that file.
I also try to cop my simulator file into the desktop. but also it not work ![Uploading Annotation 2020-06-18 030257.jpg…]()
Great work!!!! Thank you