@@abd_alim my friend, l just solved this problem right now. l think i have find the reason here. If your train/test.txt are empty, this is the rreason. you open"creating-train-and-test-txt-files.py" and edit it. Find the keyword is jpeg place. we could find only 2 jpeg words here and you edit them into "jpg" and replace this in your Google Drive. Finally, restart the colaboratory work. And your training will not quit for "608 x 608 Create 64 permanent cpu-threads ".
Those who need weight files do this, it will work: just right click the weight files link and press open link on a new window, just refresh the page until downloads(2 times refrshing worked for me)
Hey can you please share this file? 5. Download pre-trained weights for the convolutional layers (154 MB): its not available now, link is not responding Thanks a lot
Thanks so much for this course. It really helped in my ongoing final year project. Someday I might also have a video to documetmy processes too, challenges I faced and how I was able to solve them.
I found some problems in your video. 1. All images should be in JPEG format(not jpg) or ur train.txt and test.txt will be blank(according to some comments below). U might need to write a python scipt to convert them in batches. 2. 11:22 ur unzip location is wrong.
this dudes always do that. for solution of problem 1, just open the codes and change 'jpeg' to 'jpg'. for problem2, you'll figure it out if you think of
Thank you so much for explaining so clearly.. No tutorial explains in such depth on how to do especially on windows.. Thanks Sir. Please also show how to run the model on google colab too with webcam in Google colab
Thanks for the video I have a task which is pedestrian detection and congestion measurement I need to build a model that can detect pedestrian and also be able to say if the frame is congested For example let's say in a frame greater than 20 people means congested
I think: line 20: max_batches = max((number class)*2000, 6000). line 22: steps = 80%, 90% max_batches. Ex: number class = 1 --> max_batches=6000, steps=4800,5400
Thank you for making this video. 5. Download pre-trained weights for the convolutional layers (154 MB) - link is not working. Can you please share it again
anyone following along using the label studio app which is newer version of labelmg should just annotate their dataset and then press exposrt scroll down and then click export in yolo format and then follow along
@MOHAMMAD ABDUL RAZZAQ 18BCD7128 Check out this i have solve this and got a better solution colab.research.google.com/drive/1GEiDDpDOSeGqi1nijW0yCmFApQkypkmt?usp=sharing
Exactly like me and i found the solution!!! - First open de custom_Data folder and check if train.txt and test.txt are blank, if they this is the same problem! When we use "!python TFT_Data/creating-train-and-test-txt-files.py" the command expect imagens with the extencion .jpeg so others extencions dont work - Second make sure all imagens on custom_Data folder have extension .jpeg - Remove test.txt and train.txt file, return to colab and run "!python TFT_Data/creating-train-and-test-txt-files.py" - Check if the new test.txt and train.txt have content - Use the last comant of the video normaly to train YOLO
good tutorial, I appreciate your work, is it possible you can do a tutorial about precision and recall between the two models(yolov3,v4) performance and plot on the same graph
Thx for the clear explanantion. But... where is the .ipynb file u used in the video? Am I supposed to type everything? Besides ur command of 11:22 is wrong. That will extract all images in the folder of yolo_custom_model_Training rather than in the folder of custom_data.
Thank you for the great tutorial, it works! I have a questions: How do I test the model that I just trained so that I can analyse the validation accuracy of the model? Thanks
Hello, I have come to the end, the training has started. my dataset is less. After 1 hour of training, only yolov3_custom_last.weights appeared in the backup file. Where am I going wrong? Please tell me.
if you guys have a problem like "/bin/bash: darknet/darknet: Permission Denied"., after GPU-enabled. Please following this step; 1. restart your colab again and execute your whole cells again 2. if you find the error after run !make , you have to delete the obj folder inside darknet folder and then rerun !make again. 3. try to run the !darknet/darknet .... again. It works for me. if you have the problem on the train and test.txt show as the empty files, you have to edit the creating-files-data-and-name, creating-train-and-test-txt-files files by changing jpg to jpeg, or vice versa. Hope it help!
@@roy--- hey roy, I had the same problem after used the labelImg tool. It will generate class_id start from 15 to the end. The class_id should be started at 0 to the end. You can edit from your text file or using other labelImg tool.
@@srisupangth hey, when i run training command(!darknet/darknet....) CUDA status Error: file: ./src/dark_cuda.c : (). CUDA Error: no CUDA-capable device is detected: Bad file descriptor darknet: ./src/utils.c:331: error: Assertion `0' failed. i had to face with this problem. My make file value is GPU=1, CUDNN=1, OPENCV=1
I'm trying to detect objects in 4 classes and 10 images containing 100 labels in each image. But I have a performance bottleneck and the estimated time for loading images is 190 hours. What should I do to solve the problem?
sir, training stoped here // Detection layer: 82 - type = 28 Detection layer: 94 - type = 28 Detection layer: 106 - type = 28 Resizing, random_coef = 1.40 608 x 608 Create 6 permanent cpu-threads // and after this no weight file in backup folder
I had the same issue. Check out if your train.txt and text.txt are empty, in that case, open your creating-text-and-text-txt-files.py and go to the line 74. Check if the parameter is the same as the extension of your images, if not, change it. In my case looks like: "if f.endswith('.jpg'):". Hope it help you all.
@@whenmathsmeetcoding1836 yes... please make a video on how to calculate mean average precision metric. Also, thanks for sharing the knowledge for custom model. It's awesome 😀
@@whenmathsmeetcoding1836 bro great hard work. I want to ask one question. As we are just passing the images name from the train.txt and test.txt file to train the model but where we are passing the images annotations to that model?
HI ~ Teacher , I have a question when I excute the code {!make} . Content appears warning {nvcc warning : The 'compute_35', 'compute_37', 'compute_50', 'sm_35', 'sm_37' and 'sm_50'} Will it effect to excute !darknet/darknet? I excute {!darknet/darknet} appear the content /bin/bash: darknet/darknet: Not a directory how can i resolve the erroe ? thank u ~
Hello I worked with this model perfectly but the issue is the only last weight file gets save into my backup folder. Other weights file like yolo_1000 weights these files are not getting save in my backup folder. Please help me to solve this issue. Why this is happening ?
My training is abruptly stopping after 3-4 hrs. (no error message) and nothing is stored in backup folder. I am training for 2 classes. What could be the possible reasons.
great hard work. I want to ask one question. As we are just passing the images name from the train.txt and test.txt file to train the model but where we are passing the images annotations to that model?
Really helpfu!!! But what if i want to add more objects with existing yolov3 objects. I have some costomized objects but also want old one. Pls reply...
same with me, the code finished correctly displaying the network layers details which end at this lines: 608 x 608 Create 6 permanent cpu-threads Anyone can assist me?
"detections_count = 0, unique_truth_count = 0 class_id = 0, name = Wounds, ap = 0.00% (TP = 0, FP = 0) for conf_thresh = 0.25, precision = nan, recall = nan, F1-score = nan for conf_thresh = 0.25, TP = 0, FP = 0, FN = 0, average IoU = 0.00 % IoU threshold = 50 %, used Area-Under-Curve for each unique Recall mean average precision (mAP@0.50) = 0.000000, or 0.00 % Total Detection Time: 0 Seconds" I tried to train a model with 1000 images(80:20) and while detection I am getting MAP as zero...is it an issue of not detecting labels or less amount of data being trained?. The annotations in yolo format (text files) with the same name as training image are put in the same directory. Pls let me know how the model detects the annotations as I am not clear how to pass labels into in the training process.
Hello Sir, First of all, thanks for sharing such a brilliant video. Really appreciate it 🙏. Sir, I am facing the same problem of getting a error msg : "/bin/bash: darknet/darknet: Permission Denied". But the solution comman you shared isn't working in my system. Can you please help me with a solution 🙏
try this line of code it will be working >>> %cd /content/drive/MyDrive/yolo_custom_model_training >>> !darknet/darknet >>> !darknet/darknet detector train custom_data/labelled_data.data darknet/cfg/yolov3_custom.cfg custom_weight/darknet53.conv.74 -dont_show
People usually say the higher the batch, the better (try using 16, 32 or 64), subdivision should be higher deppending on your batch chosen, and memory (example, on colab I choose batch as 64, and subdivision as 8, I got an error. Then used subdivison as 16 and did it. Try using 8 or 16 for subdivison). For steps, you can try using the first value being 70% of max batches, and the second, 90% of max batches.
Does it work with multiple objects in same image? I'm trying to detect beans over a white screen. I'll try everything the video shows, but does somebody knows if there's no problem in putting more than one square over the same image using label image? Edit: IT WORKS! I've done it. Congratulations Mr. Jay, your video is very good for us people who have difficulty in learning advanced programming. I just have a simple suggestion, to add subtitles to the video. Other than this, everything is great!
thanks for the tutorial that was wonderful. when editing the MAKE file which is in the darknet folder what if i make CUDNN_HALF=0 to CUDNN_HALF=1 what is the diirence?
while running !python custom_image/creating-files-data-and-name.py i am getting below error: Traceback (most recent call last): File "custom_image/creating-files-data-and-name.py", line 58, in with open(full_path_to_images + '/' + 'classes.names', 'w') as names, \ FileNotFoundError: [Errno 2] No such file or directory: 'E:/Data Science/DL/Dataset/custom_image/classes.names' Please help
Hai, i have followed the procedure as shown in this video and everything is worked fine but when it comes to detecting an object if i show images infront of web cam its not working means not detecting the object , please let me know your response at the earliest
i am getting this error If error occurs - run training with flag: -dont_show Unable to init server: Could not connect: Connection refused can someone help me
Hi Dudi Its the common error from google colab if you reconnect form one instance and your drive is already connected with previous instance can you run and compile complete colab script again after closing it once and again restart it (RUN WITH GPU Enabled)... also sometime you need to add execute permissions like so: chmod u+x darknet/darknet (add complete command here )
Hello Sir, First of all, thanks for sharing such a brilliant video. Really appreciate it 🙏. Sir, I am facing the same problem of getting a error msg : "/bin/bash: darknet/darknet: Permission Denied". But the solution comman you shared isn't working in my system. Can you please help me with a solution 🙏
@@habibashaheen4906 Yes. My images file some photos aren't label. I delete them and problem solve. You should check your image file, maybe you missed it some photos.
Hello, want to ask. Why does my training keep on stopping midway? It always stopped after some epoch and didn't save any backups. I ran a datasets with total of 737 images to be classified into 7 different categories, and my yolo config is as follow>> batch = 32 subdiv = 8 channel = 7 max_batches = 14000 stteps = 13000, 14000 Thanks
I have followed your code but couldnt get to train the dataset. It stops at 608 x 608 Create 6 permanent cpu-threads. No further process and my backup file in google drive is empty. Can I know what is the problem with this
Exactly like me and i found the solution!!! - First open de custom_Data folder and check if train.txt and test.txt are blank, if they this is the same problem! When we use "!python TFT_Data/creating-train-and-test-txt-files.py" the command expect imagens with the extencion .jpeg so others extencions dont work - Second make sure all imagens on custom_Data folder have extension .jpeg - Remove test.txt and train.txt file, return to colab and run "!python TFT_Data/creating-train-and-test-txt-files.py" - Check if the new test.txt and train.txt have content - Use the last comant of the video normaly to train YOLO
Hello everyone, i have done everything but program did not run actually program runned and camera openned but did not detecth my custom objects i need some help abouth it. Could you please help me about this subject it's so important for me
Thank you for your video. For labelling objects in the image, can we use polygons or circles for annotating them and feed it to YOLO? If yes, can you please explain that?
The jupyter notebook is missing at the link you shared So one needs to make note of every word you are telling to be able to implement this. Please also share the link for magic command you are mentioning in the video (13th min).
definitive solution of /bin/bash: ./darknet: Permission denied just run !chmod +x ./darknet.
after running the above code !chmod +x ./darknet its showing permission denied can u plz help
chmod: invalid option -- 'd'
Try 'chmod --help' for more information.
kindly solve the errors
/bin/bash: darknet/darknet: Not a directory
@@worthtech3249 same here , do you find any solution?
[yolo] params: iou loss: mse (2), iou_norm: 0.75, obj_norm: 1.00, cls_norm: 1.00, delta_norm: 1.00, scale_x_y: 1.00
Total BFLOPS 65.879
avg_outputs = 532444
Allocate additional workspace_size = 12.46 MB
Loading weights from custom_weight/darknet53.conv.74...
seen 64, trained: 0 K-images (0 Kilo-batches_64)
Done! Loaded 75 layers from weights-file
Learning Rate: 0.001, Momentum: 0.9, Decay: 0.0005
Detection layer: 82 - type = 28
Detection layer: 94 - type = 28
Detection layer: 106 - type = 28
Resizing, random_coef = 1.40
608 x 608
Create 6 permanent cpu-threads
HEllo i am not getting trained images in backup folder.
Can some one help?
it is because the test or train.txt is empty
@Marek Žáčik I don't remember the solution of how I fixed it but you have to check the location and file name
Did you find the solution? Im stuck in the same place
@@Faisalamin01 can u plz tell me plz how to fix this?
@@abd_alim my friend, l just solved this problem right now. l think i have find the reason here. If your train/test.txt are empty, this is the rreason. you open"creating-train-and-test-txt-files.py" and edit it. Find the keyword is jpeg place. we could find only 2 jpeg words here and you edit them into "jpg" and replace this in your Google Drive. Finally, restart the colaboratory work. And your training will not quit for "608 x 608 Create 64 permanent cpu-threads ".
Those who need weight files do this, it will work: just right click the weight files link and press open link on a new window, just refresh the page until downloads(2 times refrshing worked for me)
I can't thank you enough sir !
You solved all of my doubts, thanks for giving your valuable time to us.
Hey can you please share this file?
5. Download pre-trained weights for the convolutional layers (154 MB):
its not available now, link is not responding
Thanks a lot
hey can please share me the pretrained weights ???
Thanks so much for this course. It really helped in my ongoing final year project. Someday I might also have a video to documetmy processes too, challenges I faced and how I was able to solve them.
share
hello i want to ask, are we only learning training datset only?or test and validation should learning too thanks
I found some problems in your video. 1. All images should be in JPEG format(not jpg) or ur train.txt and test.txt will be blank(according to some comments below). U might need to write a python scipt to convert them in batches. 2. 11:22 ur unzip location is wrong.
this dudes always do that. for solution of problem 1, just open the codes and change 'jpeg' to 'jpg'. for problem2, you'll figure it out if you think of
Thank you so much for this step by step tutorial video.
With the help of this video we successfully completed our Final Year Project.👌🤝🤞👍
can you share me the pretrained weights??
@@ballajaisheel1120 Sorry we done that project in 2022. Even I don't have that code😞
Nice video on yolo object detection explained with detail steps..I used same steps to run on my custom dataset.
Thanks for sharing this video.
Thank you so much for explaining so clearly.. No tutorial explains in such depth on how to do especially on windows.. Thanks Sir.
Please also show how to run the model on google colab too with webcam in Google colab
Thankyou after watching from5,6 places this is the best tutorial .
Thanks Jay. You explained it very nicely. Now, I can just get the data and start the training.
Really appreciate your efforts.
Did you run the code?
@@ahmedhusham7728 Not yet, I'm still setting up the env.
@@saitejamalladi ok, if you run it correctly please contact me via ahm179@yahoo.com
@@ahmedhusham7728 sure. Will try this weekend.
@@saitejamalladi have u tried it?
Thanks for the video
I have a task which is pedestrian detection and congestion measurement
I need to build a model that can detect pedestrian and also be able to say if the frame is congested
For example let's say in a frame greater than 20 people means congested
I think:
line 20: max_batches = max((number class)*2000, 6000).
line 22: steps = 80%, 90% max_batches.
Ex: number class = 1 --> max_batches=6000, steps=4800,5400
Thank you for making this video. 5. Download pre-trained weights for the convolutional layers (154 MB) - link is not working. Can you please share it again
same problem sir. kindly share this link again
Got the link?
Just try another browser. Chrome is blocking it.
Try copy link and open in new tab. It will automatically download the weights
Incredible job - straightforward and educational.
I am not getting any weights file in backup folder , what am I doing wrong , someone please help
sadly i can give you one like only but you deserve a million on this
great work
bro 5 th link is not working bro
Hands down crisp and clear. Thank you.
hello sir, i'm not getting your video, where you teach how to use trained weights in local(jupyter notebook) for image detection
Very comprehensive indeed!!! A big thumbs up!!!
@
Jay Bhatt Fantastic tutorial, I really appreciate it.
Thank you sooo much sir for this detailed explanation ... ☺️☺️☺️☺️
Thanks , it was helpful , Creation of this video and level of detail must have taken lots of effort.
anyone following along using the label studio app which is newer version of labelmg should just annotate their dataset and then press exposrt scroll down and then click export in yolo format and then follow along
best video ever #luvFromPakistan
Sir training does not start but having this issue
608 x 608
Create 6 permanent cpu-threads
@MOHAMMAD ABDUL RAZZAQ 18BCD7128 Check out this i have solve this and got a better solution colab.research.google.com/drive/1GEiDDpDOSeGqi1nijW0yCmFApQkypkmt?usp=sharing
@@muhammadhananasghar3102 thanks..if possible please provide the link for further also
@@namishabhasin5857 What Kind of Help You Want Any Idea????? I will provide information according to your point
@MOHAMMAD ABDUL RAZZAQ 18BCD7128 exactly like me it is ended in a few minutes with these lines:
608 x 608
Create 6 permanent cpu-threads
Exactly like me and i found the solution!!!
- First open de custom_Data folder and check if train.txt and test.txt are blank, if they this is the same problem!
When we use "!python TFT_Data/creating-train-and-test-txt-files.py" the command expect imagens with the extencion .jpeg so others extencions dont work
- Second make sure all imagens on custom_Data folder have extension .jpeg
- Remove test.txt and train.txt file, return to colab and run "!python TFT_Data/creating-train-and-test-txt-files.py"
- Check if the new test.txt and train.txt have content
- Use the last comant of the video normaly to train YOLO
I could not thank you enough for this tutorial. May Zeus bless you (I suscribe to the ancient greek religion)
good tutorial, I appreciate your work, is it possible you can do a tutorial about precision and recall between the two models(yolov3,v4) performance and plot on the same graph
sure
Sorry, did you execute this code?
Thx for the clear explanantion. But... where is the .ipynb file u used in the video? Am I supposed to type everything? Besides ur command of 11:22 is wrong. That will extract all images in the folder of yolo_custom_model_Training rather than in the folder of custom_data.
Excellent video sir,,... thank you so much... this helped me a lot...
Amazing content, thank you very much! I learned so much
Thank you for the great tutorial, it works! I have a questions: How do I test the model that I just trained so that I can analyse the validation accuracy of the model? Thanks
remove the comment in testing lines
If we are removing different comment in testing file, does that mean we will also change all the parameters has well
i didn't find any trained files in the folder backup what can be the problem !
same here
@@skandan531 would you solve your problem facing the same problem
Thank you for your knowledge. I hope i can create model someday..........
Hello, I have come to the end, the training has started. my dataset is less. After 1 hour of training, only yolov3_custom_last.weights appeared in the backup file. Where am I going wrong? Please tell me.
is this issue resolved?
if you guys have a problem like "/bin/bash: darknet/darknet: Permission Denied"., after GPU-enabled. Please following this step;
1. restart your colab again and execute your whole cells again
2. if you find the error after run !make , you have to delete the obj folder inside darknet folder and then rerun !make again.
3. try to run the !darknet/darknet .... again. It works for me.
if you have the problem on the train and test.txt show as the empty files, you have to edit the creating-files-data-and-name, creating-train-and-test-txt-files files by changing
jpg to jpeg, or vice versa.
Hope it help!
can you help me with this error when training, Wrong annotation: class_id = 15. But class_id should be [from 0 to 0], file: dataset2/60.txt
@@roy--- hey roy, I had the same problem after used the labelImg tool. It will generate class_id start from 15 to the end. The class_id should be started at 0 to the end. You can edit from your text file or using other labelImg tool.
@@srisupangth hey, when i run training command(!darknet/darknet....)
CUDA status Error: file: ./src/dark_cuda.c : (). CUDA Error: no CUDA-capable device is detected: Bad file descriptor
darknet: ./src/utils.c:331: error: Assertion `0' failed.
i had to face with this problem.
My make file value is GPU=1, CUDNN=1, OPENCV=1
I'm trying to detect objects in 4 classes and 10 images containing 100 labels in each image. But I have a performance bottleneck and the estimated time for loading images is 190 hours. What should I do to solve the problem?
You are doing an amazing work. Thank you
Sir this is urgent. I am getting "/bin/bash: darknet/darknet: Not a directory", what should I do?
@@UCky-l_seRRkuS7rHG9AJMsQ did you solve it?
sir, training stoped here
//
Detection layer: 82 - type = 28
Detection layer: 94 - type = 28
Detection layer: 106 - type = 28
Resizing, random_coef = 1.40
608 x 608
Create 6 permanent cpu-threads
//
and after this no weight file in backup folder
Hi Sauarb please recompile the makefile and check if you have. not done any mistake with config file..
have you found a solution?
@@taibtassadit910 no not yet
@@SaurabhKumar-nm6jl Hi, i have the same problem. is your train.txt empty?
I had the same issue. Check out if your train.txt and text.txt are empty, in that case, open your creating-text-and-text-txt-files.py and go to the line 74. Check if the parameter is the same as the extension of your images, if not, change it. In my case looks like: "if f.endswith('.jpg'):". Hope it help you all.
Great tutorial, can you please make a tutorial on how to evaluate trained model in terms of Precision and Recall?
Sure object detection tutorial are under development will be out soon...
+1
@@whenmathsmeetcoding1836 yes... please make a video on how to calculate mean average precision metric. Also, thanks for sharing the knowledge for custom model. It's awesome 😀
@@whenmathsmeetcoding1836 bro great hard work. I want to ask one question. As we are just passing the images name from the train.txt and test.txt file to train the model but where we are passing the images annotations to that model?
HI ~ Teacher , I have a question when I excute the code {!make} .
Content appears warning
{nvcc warning : The 'compute_35', 'compute_37', 'compute_50', 'sm_35', 'sm_37' and 'sm_50'}
Will it effect to excute !darknet/darknet?
I excute {!darknet/darknet} appear the content /bin/bash: darknet/darknet: Not a directory
how can i resolve the erroe ? thank u ~
run once again make and %cd and python script file that
is !python custom_data/creating-files-data-and-name.py
!python custom_data/creating-train-and-test-txt-files.py
Hello I worked with this model perfectly but the issue is the only last weight file gets save into my backup folder. Other weights file like yolo_1000 weights these files are not getting save in my backup folder. Please help me to solve this issue. Why this is happening ?
My training is abruptly stopping after 3-4 hrs. (no error message) and nothing is stored in backup folder. I am training for 2 classes. What could be the possible reasons.
great hard work. I want to ask one question. As we are just passing the images name from the train.txt and test.txt file to train the model but where we are passing the images annotations to that model?
I'm unable to run the make command,
showing "chmod +x *.sh " error. please help me out.
Did you resolve it? I'm stuck in the same error.
Really helpfu!!!
But what if i want to add more objects with existing yolov3 objects. I have some costomized objects but also want old one. Pls reply...
Training stops at Create 6 permanent cpu-threads Please help me @Jay Bhatt with this i tried a lot but unable to resolve.
Check for Zilong Li's answer in stackoverflow.com/questions/62296292/yolo-not-starting-to-train/66872342#66872342
@@saksheepatil5021 Means that i have just change both the file formats right
@@saksheepatil5021 thank youu bro . you're incredible
@@saksheepatil5021 Thanks, It helped a lot. Not all superhero wear capes.
@@saksheepatil5021 Line 74?
Good job! Keep up !
the weights are not saving in the backup folder. what should I do?
Thanks so much for amazing tutorial :)
Great tutorial keep doing the good work
are you separate between training test and valid when training?or only training data you learn?
My backup folder remains empty even after successfully trained
same
please i have de same problem . have solve this problem?
same with me, the code finished correctly displaying the network layers details which end at this lines:
608 x 608
Create 6 permanent cpu-threads
Anyone can assist me?
same
did anyone solve the issue?
"detections_count = 0, unique_truth_count = 0
class_id = 0, name = Wounds, ap = 0.00% (TP = 0, FP = 0)
for conf_thresh = 0.25, precision = nan, recall = nan, F1-score = nan
for conf_thresh = 0.25, TP = 0, FP = 0, FN = 0, average IoU = 0.00 %
IoU threshold = 50 %, used Area-Under-Curve for each unique Recall
mean average precision (mAP@0.50) = 0.000000, or 0.00 %
Total Detection Time: 0 Seconds"
I tried to train a model with 1000 images(80:20) and while detection I am getting MAP as zero...is it an issue of not detecting labels or less amount of data being trained?. The annotations in yolo format (text files) with the same name as training image are put in the same directory. Pls let me know how the model detects the annotations as I am not clear how to pass labels into in the training process.
when I create !python custom_data/creating-train-and-test-txt-files.py, file train and test.txt is empty, why ? please help
check the extension of your image file, if it is other than .jpg, then try to include that extension in the creating-train-and-test-txt-files.py file
Hello Sir,
First of all, thanks for sharing such a brilliant video. Really appreciate it 🙏.
Sir, I am facing the same problem of getting a error msg : "/bin/bash: darknet/darknet: Permission Denied".
But the solution comman you shared isn't working in my system.
Can you please help me with a solution 🙏
Same.
@@bub1728 same :(
Try using !chmod 755 -R darknet when you are in the yolo_custom_model_training directory. It gives the permission to execute ./darknet
@@saitejamalladi How did you do that..can i have the full path please
try this line of code it will be working
>>> %cd /content/drive/MyDrive/yolo_custom_model_training
>>> !darknet/darknet
>>> !darknet/darknet detector train custom_data/labelled_data.data darknet/cfg/yolov3_custom.cfg custom_weight/darknet53.conv.74 -dont_show
How many photos I should upload for 1 class? And for more classes,What it should be the values for batch,subdivision and step?Please hlp,Thank you.
People usually say the higher the batch, the better (try using 16, 32 or 64), subdivision should be higher deppending on your batch chosen, and memory (example, on colab I choose batch as 64, and subdivision as 8, I got an error. Then used subdivison as 16 and did it. Try using 8 or 16 for subdivison). For steps, you can try using the first value being 70% of max batches, and the second, 90% of max batches.
The link for the pre-trained weights is not working right now. Can you update it?
I'm facing this error while training, "Couldn't open file: custom_data/labelled_data.data" Is anyone who can solve this error?
Same, i got an error """ OSError: [Errno 30] Read-only file system: 'custom_data/labelled_data.data' """
Does it work with multiple objects in same image? I'm trying to detect beans over a white screen. I'll try everything the video shows, but does somebody knows if there's no problem in putting more than one square over the same image using label image?
Edit: IT WORKS! I've done it. Congratulations Mr. Jay, your video is very good for us people who have difficulty in learning advanced programming. I just have a simple suggestion, to add subtitles to the video. Other than this, everything is great!
hey how did u train it
@@kamalinichauhan4407 Just followed the video
608 x 608
Create 6 permanent cpu-threads
after running it is just stopping here so what to do plz help sir
not able to download pretrained weights in step 5.please help
Hi,
I am getting GPU isn't used and opencv isn't used....
please help..
Did you solved the problem..?
Great tutorial India Guy.
Number of input channels should be multiple of 27 but got 9 in function getMemoryShapes, getting this error, while doing inferencing. Please suggest.
thanks for the tutorial that was wonderful. when editing the MAKE file which is in the darknet folder what if i make CUDNN_HALF=0 to CUDNN_HALF=1 what is the diirence?
How model will be saved in backup folder since you didn't mention that folder anywhere in the code
while running
!python custom_image/creating-files-data-and-name.py
i am getting below error:
Traceback (most recent call last):
File "custom_image/creating-files-data-and-name.py", line 58, in
with open(full_path_to_images + '/' + 'classes.names', 'w') as names, \
FileNotFoundError: [Errno 2] No such file or directory: 'E:/Data Science/DL/Dataset/custom_image/classes.names'
Please help
I made everything but program don't detecth my custom object. Have you any opinion this situation ?
The link for Labelimg is dead. Can you please update the link and also show the other free alternatives to train a model?
Sir the model training command is showing the error 'Permission denied'. Please help
hi nipun, did you solve the problem
@@zakariaesaidi843 Yeah... I just had to do the chmod +x thing
how to solve it?
@@hegngheng2665 You just add the chmod +x and then the file name before executing the training statement
@@nipunchawla4064 ok i try later
Thxxx
hi sir i follow the all process but backup folder is empty its not saving trained data in backup folder
Same Problem here
Sir please give some example of using resnet50, alexnet and other models using for own dataset
check out letest video on channel.. its already done..
/bin/bash: darknet/darknet: No such file or directory , when complete this video, getting this error .
Hai, i have followed the procedure as shown in this video and everything is worked fine but when it comes to detecting an object if i show images infront of web cam its not working means not detecting the object , please let me know your response at the earliest
I Used !darknet/darknet but i will check backup folder he is empty. Plz give me some advice how used it ???
i am getting this error
If error occurs - run training with flag: -dont_show
Unable to init server: Could not connect: Connection refused
can someone help me
sir, I am getting a msg like this while I tried to execute train command can I know the reason-
'/bin/bash: darknet/darknet: Permission denied'
Hi Dudi Its the common error from google colab if you reconnect form one instance and your drive is already connected with previous instance can you run and compile complete colab script again after closing it once and again restart it (RUN WITH GPU Enabled)... also sometime you need to add execute permissions like so:
chmod u+x darknet/darknet (add complete command here )
Hello Sir,
First of all, thanks for sharing such a brilliant video. Really appreciate it 🙏.
Sir, I am facing the same problem of getting a error msg : "/bin/bash: darknet/darknet: Permission Denied".
But the solution comman you shared isn't working in my system.
Can you please help me with a solution 🙏
@@abhinavchoudhary2581 hey did you got solution?
@@sanket.hande9 Yes I did, I used the following command to enable permission:
!chmod +x ./darknet
@@abhinavchoudhary2581 it is giving me: "chmod: invalid option -- 'd'
Try 'chmod --help' for more information."
error: nvcc fatal : Unsupported gpu architecture 'compute_30'
Makefile:92: recipe for target 'obj/convolutional_kernels.o' failed
make: *** [obj/convolutional_kernels.o] Error 1
Sir i am using yolov7 but there are jitters(lag) in real time detection how to overcome this ?
Sir, you had forgot to tell that after making changes in the make file we have to run the
!make
command.
I am facing 6 parameters threads cpu. can you help me out?
@@techrock4499 i am having the same problem. Did you find a solution?
Hello Sir, Trained module is not geeting in backup directry after sucessfull run.
Finally i have trained my model but in backup folder there no any weight file. It’s blank
I'm getting error Unsupported gpu architecture 'compute_30' after run command !make
even i am getting the same error
did you find the solution to it?
your makefiles need to modify to GPU=1 if your colab is set to GPU.
Hi. While training the model, I get the error "Can't open label file.(This can be normal only if you use MSCOCO)", how can I solve it?
@Samet YILMAZ
i'm getting the same error did you find a solution ?
@@habibashaheen4906 Yes. My images file some photos aren't label. I delete them and problem solve. You should check your image file, maybe you missed it some photos.
Hello, want to ask. Why does my training keep on stopping midway? It always stopped after some epoch and didn't save any backups. I ran a datasets with total of 737 images to be classified into 7 different categories, and my yolo config is as follow>>
batch = 32
subdiv = 8
channel = 7
max_batches = 14000
stteps = 13000, 14000
Thanks
I am also having the same problem, no models are generated in backup folder
U find solution?!
@@akashvasu if its got stucked, you need to redo the process from the very beginning, including deleting the pasted files and folders
I have followed your code but couldnt get to train the dataset. It stops at 608 x 608
Create 6 permanent cpu-threads. No further process and my backup file in google drive is empty.
Can I know what is the problem with this
Same here. Did you find a solution?
Exactly like me and i found the solution!!!
- First open de custom_Data folder and check if train.txt and test.txt are blank, if they this is the same problem!
When we use "!python TFT_Data/creating-train-and-test-txt-files.py" the command expect imagens with the extencion .jpeg so others extencions dont work
- Second make sure all imagens on custom_Data folder have extension .jpeg
- Remove test.txt and train.txt file, return to colab and run "!python TFT_Data/creating-train-and-test-txt-files.py"
- Check if the new test.txt and train.txt have content
- Use the last comant of the video normaly to train YOLO
this was very useful for me. I appeciate this video
I am not getting any weights file in backup folder , what am I doing wrong , someone please help
For the cloning of darknet, it did not appear in any of the darknet folder that I create in the Google drive
Me too. I tried it several times and still it didn't work. So I started the whole procedure again using another Gmail account and then it worked.
Hello sir, my train.txt and test.txt showing blank data. How will i remove this error. Please help
I has the same issue and i resolved it. Ensure all your pictures are in JPEG format!
@@malekayesh6212 yeah thanks
where the weights will be saved?
@@ankitsharma1943 in the "yolov3_custom_last.names" inside your "backup" folder in Google drive
@@malekayesh6212 how can i check the accuracy of my model
where will the trained file be saved in our drive after training is completed sir?
my train.txt and test.txt ends up empty, I didn't rename my photos, they didn't have spaces, can be that the problem?
Hello everyone, i have done everything but program did not run actually program runned and camera openned but did not detecth my custom objects i need some help abouth it. Could you please help me about this subject it's so important for me
Will it be any different if I use yolov8, if there are any could you explain ?
Thank you for your video. For labelling objects in the image, can we use polygons or circles for annotating them and feed it to YOLO? If yes, can you please explain that?
No sir, standard yolo algorithm only takes a box as input
sir the script "rename_files" gives permissionerror [WinError 5] Access is denied
how to solve this kindly help me
The jupyter notebook is missing at the link you shared
So one needs to make note of every word you are telling to be able to implement this.
Please also share the link for magic command you are mentioning in the video (13th min).
GPU isn't used
OpenCV isn't used - data augmentation will be slow
have you found the solution?
what about YOLOv4-tiny? where to make changes?