First of all great tutorial :) Keep doing it. Is there any chance that you'll prepare/create a tutorial about gathering some statistics about tracked objects e.g. distance covered in total(let say in meters or km), current speed and so one. That could be really awesome to see content like this :) I hope you enjoy this idea. Best regards
Great video. I have trained my custom yolov4 model according to your previous video. Can I run deepsort tracker on it? will it be enough to replace -- model yolov4 name to my --model yolo-custom_last.weights?. Thankyou
To the guys for whom no bounding boxes appear in the video : create a new environment as he says in the beginning of the video and use this, rather than using one which you might have created earlier for some other purpose. This worked for me.
I came across your channel few days ago and have been following all videos since then, I had the exact same project that you have already executed in this video and I really appreciate your work. Keep doing these videos it's an enormous source for learning practically.
Nice video! I have a question regarding DeepSORT. DeepSORT is trained to extract features only for the person class and not for other object classes. Did you make any changes to the model or do you use the Deepsort's feature extractor trained for people on all object classes? It would be great if you could make a video on multiple object classes tracking using deepsort.
Thank you for an awesome clear explanation, on object tracking. If I convert the tiny YoloV4 model to tensorflow lite can I use Coral USB accelerator to accelerate the inferencing? I just need a way to make a post process quantized edge tpu .tflite model.
You did a very good job here. I downloaded the code, followed your instructions and it worked perfectly well on my Windows 10 machine (with gpu). There is just one strange thing. I followed a ball which is in plain sight all the time. No interruption no obstacles. Still it finds the same ball with three different tags on it (1, 2 and 4). It guess it looses the ball for some millisecs / one or more frames and finds it again in the next frame and starts a new counter. I will find out :-) Thank you very much for the excellent video!
When I try convert the model: conv_weights = conv_weights.reshape(conv_shape).transpose([2, 3, 1, 0]) ValueError: cannot reshape array of size 4554552 into shape (1024,512,3,3) What should I do?
Great video, But can someone please explain how the deepsort is working, how it is trained on the 80 classes to track them ?, or is it possible that it can track any type of object? Thanks in advance
@The AI Guy Thank you for your work. I followed your instruction, all works like in your video, but I have nor detections neither tracking. Output avi == input avi, and when I add --info in command line, I can see only FPS, no detection&tracking info.🤔
@The AI Guy . Your videos are very useful, thanks. I have one question, instead of using a webcam is it possible to acces via RTSP protocol for security cameras?
I'm trying to run the object_tracker script, but it get stuck on frame #1 and doesn't move. It seems like a problem with the model conversion to TF. Can someone guide me in the right direction?
@@TuanNguyen-gf8np bạn dùng cái của video này hay cái khác vậy? mình chạy code của (ua-cam.com/video/_zrNUzDS8Zc/v-deo.html) mà ko cần modify gì vẫn chạy đc
Hi first of all thank you so much for putting up such a nice video, could you tell if we are interested in tracking only single person instead of all. can we do it?
First of all congratulations on the great Deep-sort tutorial with Tensorflow and YOLOv4. You have explained it perfectly! I have a question: is there any way to be able to display in the demo total objects detected by class as in your yolov4-custom-functions repo? I have copied functions.py and I have imported the library, FLAGS and the script in object_tracker.py but unfortunately I can only display the counted objects in the terminal and not in the demo. I would greatly appreciate if you could help me. Thanks in advance. Regards.
Hi Ai-Guy I have a quick question. I would like to train my own data set. Today I have labeled all the pictures there, there are pictures on which no objects to be trained are visible. Do I also have to create a (picture) .txt for these pictures, which are simply completely empty? Or do Yolo and Tensorflow take it that way, if there is no text file for an image, that there is nothing to be recognized there?
I get an error of Illegal Instruction (core dumped) when trying to convert the weights into a tensorflow model using this instruction: python save_model.py --model yolov4. I am using the conda commands and so believe that I have the right version of all the packages installed. Any idea on how I could solve this error? Any help would be appreciated!
hey, I would like thank you very much for this video. If you see this comment, could you let me know if there is any way to run this directly from a python script on live video?
For tracking tho, how much images for training to give high accuracy? Because I tried using my custom weights, it doesn't give high accuracy compare to just detection system using tensorflow. Does it need to be more than 2000+ images?
One question is about the DeepSORT tracker, i think it is trained with MARS dataset(only person), what i want to confirmed is that , does it track on any objects even though the DeepSORT model is only trained on person?
CheeSiang Leow the deepSort model is indeed trained on the person class but from my experience it works extremely well on all classes. It will still track all classes, just might not be as accurate at reidentifying previously tracked objects!
Hi, In deep sort paper its mentioned that to compute Cost metric C, we combine both metric using weighted sum Lambda. In implementation how we can increase the weightage of Appearance metric to reduce Id switching issue.
Hi,thanks alot for such an awesome video.i have tried your previous video on tracking with sort and yolov3 but when I filter that for only one class I am geri getting the wrong IDs i-e getting jumps in IDs if our last id is 5 then it may jump from 5 to 13 and assign ID no 13 to the next person from ID 5.can u please suggest weather this will help me in this regard or not?
Dear Sir...Its nice watching your educative lectures in UA-cam and i have been learning alot but along the way , i encountered problem after running this python object_tracker.py --video ./data/video/test.mp4 --output ./outputs/demo.avi --model yolov4 and this what ive been getting and i tried installing and added it to my path both 32 and 64 bit but getting same errors "Could not locate zlibwapi.dll. Please make sure it is in your library path!"
Hello! Congrats for the model. It's really accurate and stable. I'd want to ask you if it is normal that the YOLO v4 algorithm takes 2 hours to process a 5 min video. I guess that something's wrong with my configuration. If it was designed to detect objects with the cam, it should be processing videos as fast as in real-time. I installed your "conda" configuration code using the tensorflow-gpu yml, though. Is it possible to speed it up? Thanks!
@Martin De Meio Hey! Its usually because you arent processing the video on a GPU. Maybe your system doesn't have a GPU, or if it does, is not compatible.
@The AI Guy could you please suggest what can be done to identify facial features of people so that I can detect whether people on the video are wearing mask or not.Please help!!!
Great work as always! I have one question. In your earlier videos when you implemented yolov3 in webcam and videos, it had around 9-10 fps, and the tiny version achieved around 30-40 fps. I am wondering why the fps for yolov4 is so low here when the paper says that yolov4 works in real-time?
@@belqesgharama347 I did integrate yolo with deepsort but could not do the evaluation. But I saw on the Github page that they were working on building an object tracker based on yolo. Maybe that's something you can evaluate using object tracking metrics.
Niyonsenga Evariste this will work in colab as long as you run the tracker command with the ‘-dont_show’ flag to hide the video output. You can then go to the location of the saved video and watch it.
Great video! Do you know if there's a way to get the mAP on this model? I can't find the code to save the detections to the file in order to run any mAP algorithms.
Is there a method to get an output based on the object detected. If a car is detected, print("car detected") I'm getting some errors doing this so if you can provide me with a simple script part, I'll add it to the main script
Nice video!! I wanted to know how can I combine object detection model and Text recognition model and run them simultaneously. Like I want recognize the characters on number plate So what i did was i used YOLOv3 for object detection and Pytesseract for character recognition. But the issue is pytesseract is not working efficiently so i wanted to create my own model for that and combine it with the object detection? Can you please Let me know the proper steps for doing so
i'm getting an error: ValueError: Memory growth cannot differ between GPU devices so where i need to do changes in code ,i've tried but every time failed to do. My system having 2 gpu's.
@@reefjames6302 YOLO's C code is pretty heavily optimised, I wouldn't expect tensorflow to be much more efficient, if at all. There might be a difference if you don't have a GPU, but in this video he says he's using a GPU.
Hi! thanks for the great videos. Do you have any github or toturial showing how to impliment obj tracking with deepsort and pytorch? Or aybe going more in depth how to pass the data back and forth between kalman filter and yolo? I would like to impliment the tracking from scrach with yolo5 and deepsort and I need to understand everything in depth.
Hey AI-Guy, I have a stupid question.(I use Yolov4 and Tensorflow 2.0) I have trained a custom dataset. By me, the directory "yolov4-416" is empty, after training. It created the file "yolov4.data-00000-of-00001", is this the weights-file? And how can i use it?
how would you do all this with opencv? i dont think i quite know what do what yet between opencv, yolo, deep sort and tensorflow, feel like is just terms that gets thrown arround but i have no idea how to isolate their uses and combination posibilities
How can I do real time object tracking on a CPU. I want to track my hand through my webcam but If I use a tracker only then it gets stuck on the face..... So I need to use detection + tracker but that gets slow..... Plz help me sir 🙏
Hi. I'd like to know if there's a way to access the specific id detections and if they can be used for computations (e.g. traffic density)? I'm only using 'car' for my classes. Thanks.
@@TheAIGuy The output is the same even when I set it to mp4. By the way, is there any way to output the results of the --info flag into a text file whenever I run it?
Hey @TheAIGuy, thanks for such a great video. I wanted your help on using this code for a custom YOLOv3 model(trained on darknet, with 3 custom classes). I cannot get the detections to show up. What do I set to make it work? Thanks!
Hey, actually I tried this out and it turned out to be too low FPS while renderring, actually I want to implement this in a real-time moving camera where I will be detecting other vehicles on the road. What might be the solution to this very low FPS that I'm getting? (I tried using tiny weights too, which was very inaccurate, yet not giving too much FPS to be real-time)
Thank you for the good lecture. Press Like!! There is a function I want to add, but please give me a lecture. traffic counting with yolo v4 and deep sort. Please tell us the part you need to edit and add in the source code.
Hey, great content, thx for the repo. Would it be possible to track an object of a class and then count it. Once it's out of the visible frame and a new object of same class is detected add it to the sum of the previous counted ones. Iam looking for smt like this but couldn't find it, except for open data cam, but I can't use that in colab.
@@TheAIGuy I would be really keen to learn to train a custom MobileNetV2-SSD, however the documentation isn't as clean as Tiny Yolo, so would be very interested in you doing a tutorial :)
Let me know what additions you would like added to the code and if you run into any issues. Cheers :)
Nice function, but the deep sort id sometimes changes within the same object, how to tackle this issue?
Farhan Tandia in the track.py file you can make the ‘age’ variable bigger. This variable is how long an ID can be not seen before it forgets it.
First of all great tutorial :) Keep doing it.
Is there any chance that you'll prepare/create a tutorial about gathering some statistics about tracked objects e.g. distance covered in total(let say in meters or km), current speed and so one.
That could be really awesome to see content like this :)
I hope you enjoy this idea.
Best regards
Great video. I have trained my custom yolov4 model according to your previous video. Can I run deepsort tracker on it? will it be enough to replace -- model yolov4 name to my --model yolo-custom_last.weights?. Thankyou
I would like to test my pre-trained weights for plate detection, how can I proceed?
To the guys for whom no bounding boxes appear in the video : create a new environment as he says in the beginning of the video and use this, rather than using one which you might have created earlier for some other purpose. This worked for me.
You have no idea how this video saved me, I cannot thank you enough
I came across your channel few days ago and have been following all videos since then, I had the exact same project that you have already executed in this video and I really appreciate your work. Keep doing these videos it's an enormous source for learning practically.
while looking for a yolov4 repository I found yours, fully documented and ready to go! What a goldmine! Works so well, many thanks!
Nice video! I have a question regarding DeepSORT. DeepSORT is trained to extract features only for the person class and not for other object classes. Did you make any changes to the model or do you use the Deepsort's feature extractor trained for people on all object classes? It would be great if you could make a video on multiple object classes tracking using deepsort.
Great video, well documented code and sturdy implementation 10/10
Thank you for an awesome clear explanation, on object tracking. If I convert the tiny YoloV4 model to tensorflow lite can I use Coral USB accelerator to accelerate the inferencing? I just need a way to make a post process quantized edge tpu .tflite model.
can you please do a video for custom object training for Deep SORT. like to train deep SORT on your data set
Hey. Were you able to figure anything out? I'm trying to train deep SORT on custom object tracking, but I'm really confused what to do.
You did a very good job here. I downloaded the code, followed your instructions and it worked perfectly well on my Windows 10 machine (with gpu). There is just one strange thing. I followed a ball which is in plain sight all the time. No interruption no obstacles. Still it finds the same ball with three different tags on it (1, 2 and 4). It guess it looses the ball for some millisecs / one or more frames and finds it again in the next frame and starts a new counter. I will find out :-)
Thank you very much for the excellent video!
When I try convert the model:
conv_weights = conv_weights.reshape(conv_shape).transpose([2, 3, 1, 0])
ValueError: cannot reshape array of size 4554552 into shape (1024,512,3,3)
What should I do?
Great video, But can someone please explain how the deepsort is working, how it is trained on the 80 classes to track them ?, or is it possible that it can track any type of object? Thanks in advance
@The AI Guy Thank you for your work. I followed your instruction, all works like in your video, but I have nor detections neither tracking. Output avi == input avi, and when I add --info in command line, I can see only FPS, no detection&tracking info.🤔
@The AI Guy . Your videos are very useful, thanks. I have one question, instead of using a webcam is it possible to acces via RTSP protocol for security cameras?
I'm trying to run the object_tracker script, but it get stuck on frame #1 and doesn't move. It seems like a problem with the model conversion to TF.
Can someone guide me in the right direction?
Question: When I run your notebook it runs well but boxes cant add to the output video!! I have same input and output (without boxes and labels :/)
If we can track objects then we can also measure speed and such right? How would you go about getting the (x,y) coordinates?
hope to see additional feature of counting the number of cars within the video and passing through the lane successfully?
Can you do this for google colab version?
EDIT: Guess what, you clearly used AI to predict my question and you answered it, thank you
Bạn chạy trên colab đã ra kết quả chưa thế ? mk thử chạy mà nó bị lỗi.
@@TuanNguyen-gf8np bạn dùng cái của video này hay cái khác vậy? mình chạy code của (ua-cam.com/video/_zrNUzDS8Zc/v-deo.html) mà ko cần modify gì vẫn chạy đc
Can you help me with something? At lines 19 and 20, I'm getting Import "tensorflow.compat.v1" could not be resolved Pylance. Please help me with this.
Hi first of all thank you so much for putting up such a nice video, could you tell if we are interested in tracking only single person instead of all. can we do it?
i folow to your instruction but the video output has no bouding box. can you tell me why?
First of all congratulations on the great Deep-sort tutorial with Tensorflow and YOLOv4. You have explained it perfectly! I have a question: is there any way to be able to display in the demo total objects detected by class as in your yolov4-custom-functions repo? I have copied functions.py and I have imported the library, FLAGS and the script in object_tracker.py but unfortunately I can only display the counted objects in the terminal and not in the demo. I would greatly appreciate if you could help me.
Thanks in advance.
Regards.
Hey Jesus, can you please explain how to counting an total object in our frame?
Hi Ai-Guy I have a quick question. I would like to train my own data set. Today I have labeled all the pictures there, there are pictures on which no objects to be trained are visible. Do I also have to create a (picture) .txt for these pictures, which are simply completely empty? Or do Yolo and Tensorflow take it that way, if there is no text file for an image, that there is nothing to be recognized there?
Martin Schmit you have to create a blank txt file i believe. Otherwise you will get an error i think.
I get an error of Illegal Instruction (core dumped) when trying to convert the weights into a tensorflow model using this instruction: python save_model.py --model yolov4. I am using the conda commands and so believe that I have the right version of all the packages installed. Any idea on how I could solve this error? Any help would be appreciated!
hey, I would like thank you very much for this video. If you see this comment, could you let me know if there is any way to run this directly from a python script on live video?
Let's focus on optimizing it, to reach realtime.
How to store those bounding box coordinates to use them as trajectory data? Please explain this also.
How do I create my own custom file. I need to do some down stream task with the tracking ID and bounding boxes.
Hi sir
I want to do this with my own data set but how to proceed
For tracking tho, how much images for training to give high accuracy? Because I tried using my custom weights, it doesn't give high accuracy compare to just detection system using tensorflow. Does it need to be more than 2000+ images?
One question is about the DeepSORT tracker, i think it is trained with MARS dataset(only person), what i want to confirmed is that , does it track on any objects even though the DeepSORT model is only trained on person?
CheeSiang Leow the deepSort model is indeed trained on the person class but from my experience it works extremely well on all classes. It will still track all classes, just might not be as accurate at reidentifying previously tracked objects!
@@TheAIGuy Thank you for reply! you solved my question keep on my head!
Hi, In deep sort paper its mentioned that to compute Cost metric C, we combine both metric using weighted sum Lambda. In implementation how we can increase the weightage of Appearance metric to reduce Id switching issue.
thanks for your extremely helpful content , I have a question is there a way to detect only one class using colab ?
Hi,thanks alot for such an awesome video.i have tried your previous video on tracking with sort and yolov3 but when I filter that for only one class I am geri getting the wrong IDs i-e getting jumps in IDs if our last id is 5 then it may jump from 5 to 13 and assign ID no 13 to the next person from ID 5.can u please suggest weather this will help me in this regard or not?
Can this be used with blue iris for monitoring security cameras?
Can you show us how to evaluate the tracking using py_motmetrics with deepsort?
Can we do this in google colab?, as my computer doesn't support gpu i am facing difficulty.
Dear Sir...Its nice watching your educative lectures in UA-cam and i have been learning alot but along the way ,
i encountered problem after running this
python object_tracker.py --video ./data/video/test.mp4 --output ./outputs/demo.avi --model yolov4
and this what ive been getting and i tried installing and added it to my path both 32 and 64 bit but getting same errors
"Could not locate zlibwapi.dll. Please make sure it is in your library path!"
Hello! Congrats for the model. It's really accurate and stable. I'd want to ask you if it is normal that the YOLO v4 algorithm takes 2 hours to process a 5 min video. I guess that something's wrong with my configuration. If it was designed to detect objects with the cam, it should be processing videos as fast as in real-time. I installed your "conda" configuration code using the tensorflow-gpu yml, though. Is it possible to speed it up? Thanks!
@Martin De Meio Hey! Its usually because you arent processing the video on a GPU. Maybe your system doesn't have a GPU, or if it does, is not compatible.
how to use the code without Cuda or does it specifically have to use Cuda to run the code?
@The AI Guy could you please suggest what can be done to identify facial features of people so that I can detect whether people on the video are wearing mask or not.Please help!!!
This can be done by training custom data set.. kindly refer his previous videos.
What about same person or car come again, will count new for it
Good Video! Can you make in the future a stock price prediction model?
Will do for sure!
Can you tell me, about the CPU usage while using deepsort ? for me it was always close to 100, is there anything that i'm doing wrong ?
Hey, the google drive link has expired for the weights. Is there an updated link anywhere?
Great work as always! I have one question. In your earlier videos when you implemented yolov3 in webcam and videos, it had around 9-10 fps, and the tiny version achieved around 30-40 fps. I am wondering why the fps for yolov4 is so low here when the paper says that yolov4 works in real-time?
I have the same question, but maybe coz i am using CPU not GPU, what are u using?
I ran it on a GPU and as far as I can recall, I was getting around 30fps (on average) on random videos.
@@zunayedmahmud3515 I tried using my GPU today it’s around 7 and 9 fbs.
By the way , did you try to evaluate yolo detector with deepsort using any evaluations codes?
@@belqesgharama347 I did integrate yolo with deepsort but could not do the evaluation. But I saw on the Github page that they were working on building an object tracker based on yolo. Maybe that's something you can evaluate using object tracking metrics.
finaly the video I was waiting for, one more question can this work perfect on google colab?
Niyonsenga Evariste this will work in colab as long as you run the tracker command with the ‘-dont_show’ flag to hide the video output. You can then go to the location of the saved video and watch it.
@@TheAIGuy Thanks so much, I will work on it
@@houston3414 Have you worked on it? If so I want the notebook :P
@@mohammadmoaddi2268 not yet Sir, by the way I remember The AI Guy said he will do a video about it soon
Nice video
I want to add a congestion measurement to this, do you know how please?
hey can i add a face recognition system which will recognise and name it and track the person
How would I select the bounding box of the car than inside that detection select the license plate of different colors?
Great video!
Do you know if there's a way to get the mAP on this model? I can't find the code to save the detections to the file in order to run any mAP algorithms.
Do you have any tutorial or video on how to train your yolo v4 + deepsort on your custom dataset and labels? Thanks
Superb video thank you for video it will help us a lot Sir can you tell me your system configuration
Hii, very nice tutorial. I was wondering if you could count the vehicles after tracking them.
Great tutorial! can you please tell which GPU laptop you are using?
Is there a method to get an output based on the object detected. If a car is detected, print("car detected")
I'm getting some errors doing this so if you can provide me with a simple script part, I'll add it to the main script
Hey great work with this, can we use this for lidar data aswell?
Nice video!! I wanted to know how can I combine object detection model and Text recognition model and run them simultaneously. Like I want recognize the characters on number plate So what i did was i used YOLOv3 for object detection and Pytesseract for character recognition. But the issue is pytesseract is not working efficiently so i wanted to create my own model for that and combine it with the object detection? Can you please Let me know the proper steps for doing so
Tiktok India i will be showing this in an upcoming video so stay tuned!!
@@TheAIGuy Thankyou so much. I will be waiting for that video!
Thank you so so so much! You helped me a lot :)
i'm getting an error:
ValueError: Memory growth cannot differ between GPU devices
so where i need to do changes in code ,i've tried but every time failed to do. My system having 2 gpu's.
what is the reason that you use tensorflow but not darknet?
Yeah, I was wandering too
TensorFlow, tflite, TensorRT are great for edge devices
@@GOBish23 could you be more specific plz ?
@@saifeddinemahmoudi1138 edge devices have limited hardware hence running these less intense framework is more betterer
@@reefjames6302 YOLO's C code is pretty heavily optimised, I wouldn't expect tensorflow to be much more efficient, if at all. There might be a difference if you don't have a GPU, but in this video he says he's using a GPU.
Hi! thanks for the great videos. Do you have any github or toturial showing how to impliment obj tracking with deepsort and pytorch? Or aybe going more in depth how to pass the data back and forth between kalman filter and yolo? I would like to impliment the tracking from scrach with yolo5 and deepsort and I need to understand everything in depth.
Thanks for uploading this tutorial. I will have many questions to apply this on my custom model. May I know what's the GPU that you are using?
Joko susilo i have a nvidia GTX 1060
Thanks for a really nice video! IWould it be possible for you to include rtsp streams, and not only video files or webcam?
Agreed would also like to know as my current project requires RTSP stream.
Thank you for your tutorials!! I have a question how i can import main method in another sript?
Hey AI-Guy, I have a stupid question.(I use Yolov4 and Tensorflow 2.0) I have trained a custom dataset.
By me, the directory "yolov4-416" is empty, after training. It created the file "yolov4.data-00000-of-00001", is this the weights-file? And how can i use it?
Thank you so much for this video. Btw, can I apply this code into raspberry pi with camera installation?
how would you do all this with opencv? i dont think i quite know what do what yet between opencv, yolo, deep sort and tensorflow, feel like is just terms that gets thrown arround but i have no idea how to isolate their uses and combination posibilities
is it possible to combine this as a tflite model that takes stream of images and returns boxes coordinates and id (from the tracker)
How can I do real time object tracking on a CPU. I want to track my hand through my webcam but If I use a tracker only then it gets stuck on the face..... So I need to use detection + tracker but that gets slow..... Plz help me sir 🙏
Is it possible to get the info saved into a txt or json file like the cloud tutorial does? Awesome work, keep it up :)
Hi. I'd like to know if there's a way to access the specific id detections and if they can be used for computations (e.g. traffic density)? I'm only using 'car' for my classes. Thanks.
Hi, have you gotten solution am interested too
Hello all, need help , how may i run this object tracking on webcam in colab. Regards
Hi!
@The AI Guy
Will be a tutorial in order to try this on colab?. Thanks
Harry yes currently rworking on a video to run this in colab! So dont worry it will be out shortly
@@TheAIGuy thanks man.
It would be super if you could add the ROI function. I mean that you could count objects just in a selected area.
Hey viiger, i have same question like that.Are you find it.?
you could try modifying the object_tracker file
I needed the same thing and that worked.
Can you put a tutorial on Human pose estimation please??
is it possible to use with RTSP live stream camera to dedicate the live view in real-time?
Is there a way for the output video have a normal frame rate? The output video that I always get is at 600fps which is too high.
Christian try changing the output video type to non avi file.
@@TheAIGuy The output is the same even when I set it to mp4.
By the way, is there any way to output the results of the --info flag into a text file whenever I run it?
Can I implement it with YOLOv5 with custom classes?
how to run this on multiple GPU's as I have two gpu and this code is giving me error kindly help
The Ai guy i am big fan of your vedio
I have a dought how do i calculate the accuracy of detected objects ?can you suggest some ideas
Thats very cool, how can i get this to track soccer players on my video? and could it track data? Thx
Hey @TheAIGuy, thanks for such a great video. I wanted your help on using this code for a custom YOLOv3 model(trained on darknet, with 3 custom classes). I cannot get the detections to show up. What do I set to make it work? Thanks!
I have the same issue. did you figure out?
Hey, actually I tried this out and it turned out to be too low FPS while renderring, actually I want to implement this in a real-time moving camera where I will be detecting other vehicles on the road. What might be the solution to this very low FPS that I'm getting? (I tried using tiny weights too, which was very inaccurate, yet not giving too much FPS to be real-time)
That was very helpful! I know it prints FPS at each frame. Is there an easy way to print the actual frame number from the original video?
use cv2.putText function
Hello I containerized the project, their is one error I created an issue on github
I have a trained weights of YOLOv3 can I use same weight for YOLOv4?
Hitesh Jaware no unfortunately not as the architecture for both models would be different.
@@TheAIGuy Okay thanks!
I am having 0.2 fps and 2 fps on yolo tiny. Any suggestions to increase the fps?
please do a absolute newbie on object detection workflow
Is it possible to use the anaconda route for AMD gpus?
Hello Mr AI Guy, how can i intall MINGW64 TO CLONE THE GITHUB??
You should be able to just google "install mingw64" and look at the top one or two responses.
I am getting below error:
Frame #: 1
FPS: 0.04
: cannot connect to X server
Any idea on how to sole this?
me too ! did you figure out how to solve it
comment the cv2 imshow and destroy all
Thank you for the good lecture.
Press Like!!
There is a function I want to add, but please give me a lecture.
traffic counting with yolo v4 and deep sort.
Please tell us the part you need to edit and add in the source code.
이종선 you can just copy the counting function from my previous video and add it in to object tracker script!
@@TheAIGuy We need a lecture that draws a line on a frame and counts when an object passes through that line.
@@구를베베 Hello, did you manage to draw a line to count object? I am also trying to do that but I don't know how.
@@ssatriyaa I am also struggling.Have you find it
Thanx mate
@The AI Guy can you tell how to add a new class, new object in that 80 classes trained models,our own new type of class..
you would need to custom train your data on a separate class, and then the process is the same.
@@palaache do you have any idea on how to train your custom data with yolo v4 deepsort?
Hey, great content, thx for the repo. Would it be possible to track an object of a class and then count it. Once it's out of the visible frame and a new object of same class is detected add it to the sum of the previous counted ones. Iam looking for smt like this but couldn't find it, except for open data cam, but I can't use that in colab.
Hey Tim, i have also same question .Have you find it?
good work brother
What is Deep SORT?
Are you planning on training MobileNetv2-SSD anytime?
Grace Cocks not currently planning on training mobilenet. But let me know if you want me to and i can look into it!
@@TheAIGuy I would be really keen to learn to train a custom MobileNetV2-SSD, however the documentation isn't as clean as Tiny Yolo, so would be very interested in you doing a tutorial :)