I forgot to download the chart.png file while training. Now runtime got disconnected and i dont have it now. So is there any way to get chart.png file?
Any chance you could add in your training video (or even a seperate one) how to reduce the number of items being detected? I ran yolov4 on a test video and it was discovering things that I dont care about. Thanks
Love your videos, extremely informative! I've been using experimenting with modifying the darknet_video.py file so that I can make new train data from a video instead of manually annotating many frames. Cant wait to try with Yolov4!
Hey Jeff, that sounds like a great idea. But how does one annotate over a video rather than frames, I've been thinking of switching to this method as well.
you are extremely up to date man, I'm learning AI on your channel hope this channel will get more subscribers, can you do tracking with YOLOv4 and DeepSORT in the future?
Yes it is very similar, almost the same. Refer to my old yolov3 cloud video and the alexeyab/darknet repository if you want to get a head start on training!
Hi, I'm having a problem when training videos, I even used your test.mp4, every time I upload the video to detect, it runs, but then after about 3 - 5 seconds it keep on saying "Video-stream stopped!", what could be wrong?
Really waiting for this one.. COMING SOON: HOW TO TRAIN YOUR OWN YOLOv4 CUSTOM OBJECT DETECTOR.. I hope to see this one come out soon.. Worth all 27 minutes and 40 seconds..
Yes will definitely do a video on this, I know it can be a pain to do locally so I will try and create an easy to follow tutorial for local darknet build.
I discovered the error, you have to go into the folder Darknet -> src -> and you can visualise you video.mp4 there -> copy your video folder saved in src and substitute the test.mp4 in the code with you copy folder and run it
Great video!!! I wanted to use pre-trained weights of yolov4 or yolov3 to detect only pedestrians and not other 79 objects. how do I limit the detector for only one class and also using the pre-trained model? How to implement object detection on live footage coming from cctv camera??
Hi bro , thanks so much for ur incredible tutorials, about the next video of training our own customize object detection i hope that u will include test and validation and not just train as the previous video of yolo v3 ,
Thanks for the excellent YOLOv4 video, I would like to train the detector with custom objects. How could I do it ?, Will the same procedure be used for YOLOv3 training?
Hi there, I will have a video out on training custom yolov4 model soon. It is fairly similar to how it is done in my yolov3 in the cloud video. Refer to that video and the Alexeyab/darknet github repository if you want to get a head start!
Looking into this for future videos. You can still write functions with opencv to write on images using coordinates. It is just the native way that darknet attempts to open images that does not work, opencv still works in colab as part of functions you build.
Hi, firstly thank you so much for this video. I am learning a lot of things with you :) I have a question. How to count the detected objects on the image? I should do this.
Thanks for the kind words! Counting the number of objects doesn't come as a built in flag for yolov4. However if you save results to a txt or JSON file, you could create a function to read from the file and extract number of objects!
I will definitely make a similar video for flask apis similar to yolov4. It will be slightly different as yolov4 weights don't convert the same way as yolov3 weights. There are more deep learning layers in yolov4 as there are a lot of new features added to improve results.
Great video! Is there a way to get the boxes' coordinates as a usable output and not as a text in the output itself? I would like to use it for object tracking and so on. Thanks!
Hi! Great video as always! I was wondering if you could add in how to get the performance metrics of the model (e.g mAP, Recall, Precision etc). I've been searching the net for days and can't seem to find a solution. looking forward to the next video!
Hello Sir, I have learnt use of YOLO by watching your videos. Thank you for providing such videos. Sir can you please provide us video on how to distinguish between static and moving object using YOLO in google colab?
Good job! Do you noticed more random "noise" bboxes compared to yolov3? I've compared the detections of yolov3 and yolov4 on the same video (oxford town centre) and I've noticed that yolov4 frequently show false positive. Can be possible that it needs an higher value for the threshold?
Love your Content.💪 AI I noticed w/ Colab some days I can run " darknet/data" in a command w/ no issues other times it will throw an error until I specify the full ./content/darknet/data , What's going on?
Hey! Really newbie to deep learning here. I want to train a model to detect a specific object on a video/image. Which video do you recommend to start? Thank you for your content.
Great job! I have a quesiton: What if i want to identify ships , airplanes ,cars from these huge satellite images. What would the best approach be? should i use Yolt instead of Yolo? any guides on this? Many thanks in advance!!
Thanks for the video. I ran the street inferencing example and it took 699 mS, as opposed to 32mS in your example. I had a T4 GPU in my run. Can you please explain the discrepancy? Thank you.
Thanks for such a detailed video on Yolov4. I just have one issue that I am encountering with the colab notebook. When I run the yolov4 on the 'test' video I am getting an average fps of 7.0 from the beginning till the end even though the model detects all the objects with pretty decent accuracy. Can you give me an idea why the fps is so low?
Hello~! I have tried on YoloV3 google collab build and train object detector and it works wonderfully, thank you! Here's a question tho, can Google Collab be used to make a real-time face detection system using webcam? If yes, could you give me some tips on it bcs I'm really newb in this. Thank you!
you can access your webcam through Colab, however it can't be real-time. You can only save a webcam image or video and then run it through the model after.
@@TheAIGuy oohh, thank you very much! Since my computer didn't have much strong GPU, I used Google Collab. I asked someone abt this since Google Colab can't make real time, they said to try training the data in Google Collab and test it in computer. Idk how it will work yet ;; haha
Hi. thank you for amazing video. I have one question: How could I omit bounding box labels in images?? I want to see just bounding box without labels in images.
Hi ! Can we not run webcam feed on this collab? Since it can analyze video footage do you think it works with webcam? Thank you very much for this video
Its very nice video, I am very eager to know HOW TO TRAIN OWN YOLOv4 CUSTOM OBJECT DETECTOR on google colab, as high end GPU is issue for me. it takes hours to complete training.
This awesome, excellent, but I have one issue when I want to detect video from my google drive it showed me this message(/bin/bash: ./darknet: No such file or directory), So how can I fix it?
Hey Bro how can I Output Bounding Box Coordinates in a video streaming ? In other words i want know the coordinate of de center point of the BBox and i want to watch this center point in the video streaming with his coordinates
Hello sir, I have error in this step "# run darknet detection on test images" CUDA status Error: file: ./src/dark_cuda.c : () : line: 38 : build time: Nov 22 2022 - 08:53:44 CUDA Error: no CUDA-capable device is detected Darknet error location: ./src/dark_cuda.c, check_error, line #69 CUDA Error: no CUDA-capable device is detected: Bad file descriptor how to solve this error? can you explain me, please?
hiii, nice video! I want to train my own created dataset in google colab using pre-trained model, can you kindly do a video on how to do it. thank you :)
Hey! I created a digit classifier using your tutorials. How can I use the numbers detected on a separate program? For example if it detected a 6 and 5.. How can I get my program to add those detected numbers?
yeah same case with me i cant detect small object, maybe can you help us ?, how to change backbone, and what backbone is the best for tiny object? how about tiny yolo ?
How can I see the distance between people, if I only detect the object people, such as a row of people and know their distance between people thanks for your help
Let me know if you run into any issues and which videos you would like to see next! :)
i follwed ur yolov3 custom object detection
I forgot to download the chart.png file while training. Now runtime got disconnected and i dont have it now. So is there any way to get chart.png file?
Any chance you could add in your training video (or even a seperate one) how to reduce the number of items being detected? I ran yolov4 on a test video and it was discovering things that I dont care about. Thanks
@@RobertTowster Definitely can cover this for you in an upcoming video.
Sir can you make evaulation report with yolo? Like csv or something
Thank you for such a timely video on Yolo v4. Can't wait to see the Yolo v4 Custom Object Detector video!
Hey dude, I need to do lane detection too.
have you got some idea? How to achieve that?
Excellent work with YOLOv4 in Colab 😊👍
Thanks a lot! Really enjoying all your YOLOv4 videos! Keep it up!
@@TheAIGuy Thanks bro :D. really appreciate it. Same to you.
Tell me would you be interested in a collaboration?
@@Augmented_AI I would love to collaborate! Feel free to email me as I'd love to hear more. Email is in my 'About' section.
Love your videos, extremely informative! I've been using experimenting with modifying the darknet_video.py file so that I can make new train data from a video instead of manually annotating many frames. Cant wait to try with Yolov4!
Jeff Dryden Thanks for the kind words! Appreciate it.
Hey Jeff, that sounds like a great idea. But how does one annotate over a video rather than frames, I've been thinking of switching to this method as well.
Perfect video! I'm waiting for yolov4 custom object detector on cloud.
Brilliant tutorial , the precooked notepad really helped.
A very comprehensive tutorial for a beginner like me, thanks for the effort =)
Mate, you do really great stuff! It helps dramatically. Thanks a lot. Custom image training with Colab is also awesome. Works great for me.
Waiting for the training section asap!
when ?
Your videos are so helpful and extremely informative!! thank you for this content :D
Thanks for the kind words! Wishing you all the best!
Can't wait to see the Yolo v4 Custom Object Detector video!
You are awesome. Thank you for going through the wiki and making this video.
Hey dude, I need to do lane detection too.
have you got some idea? How to achieve that?
my favourite class yolo thank you so much
really cool dude, keep it up ! looking forward for your next video
Hey dude, I need to do lane detection too.
have you got some idea? How to achieve that?
Amazing! Thanks for uploading. Can't wait for the video on the custom YOLOv4 object detector! Subscribed!!
Great video. Got more info than I hoped for.
Thank you for providing this great tutorial!
Thanks bro
you are extremely up to date man, I'm learning AI on your channel hope this channel will get more subscribers,
can you do tracking with YOLOv4 and DeepSORT in the future?
Been waiting for this!!! I hope the procedure is same a YOLOV3 for custom object detection!?
Yes it is very similar, almost the same. Refer to my old yolov3 cloud video and the alexeyab/darknet repository if you want to get a head start on training!
The AI Guy sure! Thankyou very much for your awesome videos!! Love it ❤️❤️
Hi, I'm having a problem when training videos, I even used your test.mp4, every time I upload the video to detect, it runs, but then after about 3 - 5 seconds it keep on saying "Video-stream stopped!", what could be wrong?
Big up from Russia! Красавчик 🙌
It's very helpful! Thanks a lot😁
I just spent all day getting this to run on my gpu....time to get this going on colab
Perfect Video
yeah
lol he is just copying the git hub repo of AlexyAB ( the original repo created for YOLO v4 )
Awesome Video! Thank you very much
Anschel Homero Julio Roman Wow thanks for the kind words! Glad you enjoyed.
Thanks, it's a nice video. Keep it that way
Really waiting for this one..
COMING SOON: HOW TO TRAIN YOUR OWN YOLOv4 CUSTOM OBJECT DETECTOR..
I hope to see this one come out soon..
Worth all 27 minutes and 40 seconds..
Same here
Same here
Waiting for a great tutorial, please make it as soon as possible I have a project I want to finish :(
Video is up now! Hope you enjoy!
How can I count the total number of objects?
Now u actually are a ai guy..u got a robo voice too,🖤💯😂
Hey dude, I need to do lane detection too.
have you got some idea? How to achieve that?
Legend! Could you also show its installation and training locally? Appreciate it though, thanks for the share!
Yes will definitely do a video on this, I know it can be a pain to do locally so I will try and create an easy to follow tutorial for local darknet build.
@@TheAIGuy Would be awesome, waiting for that tutorial to share with my friends also asap!
@@woosal-kc7tr Augmented Stertups has a well explained guide video on Yolov4 locally, seamless installation with GPU and execution.
@@voel.m323 you are right, but not for the local training as the moment.
can't wait for the training on custom dataset tutorial
thank you very much !! It helps me a lot!!!
Hy! thanks for the work. Why it appears the error "Video-stream stopped!" multiple times when I try to run the video.mp4?
I discovered the error, you have to go into the folder Darknet -> src -> and you can visualise you video.mp4 there -> copy your video folder saved in src and substitute the test.mp4 in the code with you copy folder and run it
@@andrepinto1901 Thanks Andre! It worked for me.
thank you for this great video^_^
great video..really very helpful.....thanks a lot...
thank you very much for sharing. beer is on me
I hope the video: HOW TO TRAIN YOUR OWN YOLOv4 CUSTOM OBJECT DETECTOR is nearly finisched because the whole topic is extremly interesting
It will be out in the next couple days! Stay tuned!
Great one :) May I know when will you launch the next video on Custom training?
Great video!!! I wanted to use pre-trained weights of yolov4 or yolov3 to detect only pedestrians and not other 79 objects. how do I limit the detector for only one class and also using the pre-trained model?
How to implement object detection on live footage coming from cctv camera??
I can do it
just replace dont_show in place of all other classes in data/coco.names and run it
@@sujitha3335 yes but doing that it outputs dont_show for the classes that are renamed
Hi bro , thanks so much for ur incredible tutorials, about the next video of training our own customize object detection i hope that u will include test and validation and not just train as the previous video of yolo v3 ,
Thanks for the excellent YOLOv4 video, I would like to train the detector with custom objects. How could I do it ?, Will the same procedure be used for YOLOv3 training?
Hi there, I will have a video out on training custom yolov4 model soon. It is fairly similar to how it is done in my yolov3 in the cloud video. Refer to that video and the Alexeyab/darknet github repository if you want to get a head start!
So helpfull.. But how can i draw some diffeeent shape using the co ordinates as there is no option to work with open cv here ?
Looking into this for future videos. You can still write functions with opencv to write on images using coordinates. It is just the native way that darknet attempts to open images that does not work, opencv still works in colab as part of functions you build.
Hi, firstly thank you so much for this video. I am learning a lot of things with you :) I have a question. How to count the detected objects on the image? I should do this.
Thanks for the kind words! Counting the number of objects doesn't come as a built in flag for yolov4. However if you save results to a txt or JSON file, you could create a function to read from the file and extract number of objects!
@@TheAIGuy This is a good idea! Thanks
thank you it is realy helpfully
its soo clear thanks
when will you release the next video on YOLOv4 training the custom dataset? I am eagerly waiting !
Deepak sahoo should be out either tomorrow or Friday!
@@TheAIGuy thank you very much!
@@TheAIGuy still waiting
Please make a video on yolov4 flask api. I urgently need this. Are the steps same as yolo v3 api? If I just copy the yolov4 weights then will it work?
I will definitely make a similar video for flask apis similar to yolov4. It will be slightly different as yolov4 weights don't convert the same way as yolov3 weights. There are more deep learning layers in yolov4 as there are a lot of new features added to improve results.
Great work 👍🏻
wow thank you so much genius!!!!
Great video! Is there a way to get the boxes' coordinates as a usable output and not as a text in the output itself? I would like to use it for object tracking and so on. Thanks!
Hi, I'm in desperate need for you to provide us the custom training. When will it come out!!????
Hi!
Great video as always!
I was wondering if you could add in how to get the performance metrics of the model (e.g mAP, Recall, Precision etc). I've been searching the net for days and can't seem to find a solution.
looking forward to the next video!
Hello Sir, I have learnt use of YOLO by watching your videos. Thank you for providing such videos. Sir can you please provide us video on how to distinguish between static and moving object using YOLO in google colab?
Done! Loaded 137 layers from weights-file
Learning Rate: 0.001, Momentum: 0.949, Decay: 0.0005
Detection layer: 139 - type = 28
Detection layer: 150 - type = 28
Detection layer: 161 - type = 28
Resizing, random_coef = 1.40
608 x 608
Create 6 permanent cpu-threads
While training Code not running after this pls tell solution
Good job! Do you noticed more random "noise" bboxes compared to yolov3? I've compared the detections of yolov3 and yolov4 on the same video (oxford town centre) and I've noticed that yolov4 frequently show false positive. Can be possible that it needs an higher value for the threshold?
thanks very much
this is perfect! thank you so much sir! I have a question, how to use the weights file in gui?
Awesome, can you please let me know if you can make real time objective detection...not local though... but in colab itself
Love your Content.💪 AI I noticed w/ Colab some days I can run " darknet/data" in a command w/ no issues other times it will throw an error until I specify the full ./content/darknet/data , What's going on?
Wonderful. Can you put a tutorial on custom object detection in yolo v4?
I am also waiting for your tutorial custom object detection with YOLO v4 on the cloud.
Hey! Really newbie to deep learning here. I want to train a model to detect a specific object on a video/image. Which video do you recommend to start? Thank you for your content.
Start on my YOLOv3 in the cloud video, it will show you how to train easily in the cloud.
How can i run custom detection on video?
Please can you give the command.
Great job! I have a quesiton: What if i want to identify ships , airplanes ,cars from these huge satellite images. What would the best approach be? should i use Yolt instead of Yolo? any guides on this? Many thanks in advance!!
Amazing tutorial !
is there anyway to run it on a live stream from webcam?
Thanks for the video. I ran the street inferencing example and it took 699 mS, as opposed to 32mS in your example. I had a T4 GPU in my run. Can you please explain the discrepancy? Thank you.
Thanks for such a detailed video on Yolov4. I just have one issue that I am encountering with the colab notebook. When I run the yolov4 on the 'test' video I am getting an average fps of 7.0 from the beginning till the end even though the model detects all the objects with pretty decent accuracy. Can you give me an idea why the fps is so low?
I got around 15.5 avg fps.
I am wondering how to use yolo4 and OpenCV on colab with GPU support.
That is what this video covers... it is already with OpenCV and GPU.
@@TheAIGuy absolutely NO. This is how to use darknet script ... no OpenCV and dnn module.
@@RemekKinas darknet is built with both Opencv and GPU enabled ...
Thanks for the clear explanation. Can you also do how to implement YOLO v4 on Azure Machine Learning Studio ? Thank you
Nikki Will definitely keep this as a future video idea. Thanks for the suggestion!
Will the changes you made in darknet effect yolov3 model of yours? That you showed before
please post soon the video for custom classifier
Hello~! I have tried on YoloV3 google collab build and train object detector and it works wonderfully, thank you! Here's a question tho, can Google Collab be used to make a real-time face detection system using webcam? If yes, could you give me some tips on it bcs I'm really newb in this. Thank you!
you can access your webcam through Colab, however it can't be real-time. You can only save a webcam image or video and then run it through the model after.
@@TheAIGuy oohh, thank you very much! Since my computer didn't have much strong GPU, I used Google Collab. I asked someone abt this since Google Colab can't make real time, they said to try training the data in Google Collab and test it in computer. Idk how it will work yet ;; haha
Hi. thank you for amazing video. I have one question: How could I omit bounding box labels in images?? I want to see just bounding box without labels in images.
You would have to use the coordinates in the txt file or JSON and write a separate function to draw just a normal bounding box without labels.
YOLOv4 seems pretty fast and accurate but needs a lot of recourses you guys should try different objects detectors
Hi ! Can we not run webcam feed on this collab? Since it can analyze video footage do you think it works with webcam? Thank you very much for this video
very well explained. could you please give me instruction or code how can i blur detected images?
Its very nice video, I am very eager to know HOW TO TRAIN OWN YOLOv4 CUSTOM OBJECT DETECTOR on google colab, as high end GPU is issue for me. it takes hours to complete training.
Hi, its an interesting video. Is there a way to grab the contour of the object detected?
hello excellent video ... how could i run it with my pc's webcam in real time
Hi Could you make a video on multi camera multi target object tracking
This awesome, excellent, but I have one issue when I want to detect video from my google drive it showed me this message(/bin/bash: ./darknet: No such file or directory), So how can I fix it?
did you fix it sir :( help me please
thank you for the video but i want to know how to not detect some objects that i don't need?
Hey Bro how can I Output Bounding Box Coordinates
in a video streaming ? In other words i want know the coordinate of de center point of the BBox and i want to watch this center point in the video streaming with his coordinates
Hello sir, I have error in this step "# run darknet detection on test images"
CUDA status Error: file: ./src/dark_cuda.c : () : line: 38 : build time: Nov 22 2022 - 08:53:44
CUDA Error: no CUDA-capable device is detected
Darknet error location: ./src/dark_cuda.c, check_error, line #69
CUDA Error: no CUDA-capable device is detected: Bad file descriptor
how to solve this error? can you explain me, please?
hiii, nice video! I want to train my own created dataset in google colab using pre-trained model, can you kindly do a video on how to do it. thank you :)
Hey dude, I need to do lane detection too.
have you got some idea? How to achieve that?
Hey! I created a digit classifier using your tutorials. How can I use the numbers detected on a separate program? For example if it detected a 6 and 5.. How can I get my program to add those detected numbers?
Hello The AI Guy, please Can you make a video on how to train a custom object detection using yolov7?
yeah same case with me i cant detect small object, maybe can you help us ?, how to change backbone, and what backbone is the best for tiny object? how about tiny yolo ?
Hello is there a possibility to make a change to show live webcam in grayscale? :D
How can I see the distance between people, if I only detect the object people, such as a row of people and know their distance between people
thanks for your help
How can I create a dataset from COCO like you did?
thank you so much.... and can you please make a video with the tflite model of this yolov4 for android ?
nice !!
Hello sir, can you tell me if there is a video or guide on implementing yolo v4 object detection on raspberry pi? Please help
Can you create video running yolov4 on multiple webcams?
Hi, is there a way to save the detections from a video as several images containing the cropped detections?
Unfortunately this is not a current flag supported by the darknet repository. I would love to see this functionality as well!
Great video? It would be great if you can create a tracking video using YOLOv4 in conjunction with deep SORT.
How can we use colab to train and then download the trained model to use locally?
Brent Edds check out my latest video for training model in Colab! Once its trained just download the model’s weights file to your local machine!
Unable to init server: Could not connect: Connection refused
getting this error