YOLO v7 + SORT Object Tracking | Windows & Linux

[@JulesElmurib]

is there a list somehwere of all the flags and options? TYSM this video was amazing

[@chaimacham9414]

please how can i show confidence with the id in object trucking in the bounding boxe?

[@ujjwalkatiyar9998]

Your videos helped me a lot in college project.

[@AbdullahJirjees]

Thank you for this nice video, I have a question, how can I add object counter based on the classes that I choose?

[@quoctienpham1903]

Great videoooo. How can I change the resolution of the result video because from your tutorial it wasnt display the whole video. Thank you!

[@julianelmurib6684]

Hihi, not sure why but my outputs are no longer saving, it creates a new folder in the runs folder but doesn''t save any ooutpout to it, is there a way to manually save / is there somewehre that explains all of the tags like --save-conf

[@JulesElmurib]

why is my FPS so slow is this an issue of my cpu ?

[@soumyajitpodder3516]

Hi, nice work. Can you try instance segmentation (Mask R-CNN) and object tracking at the same time ?

[@ThePontino]

Nice work guy! is there any easy way to export the tracks into a numerical format? E.g. a list of each object and its (inner) list of subsequent positions (e.g. given by bitmap x-y position)?

[@ammar46]

Hey, I have GPU, but half flag never works, so usually I keep it as false. This time the FPS is 0. Can I know how to solve it??

[@livn-vr6223]

Great video is it possible to use the sorter for multiple cameras

[@ammar46]

What should I do remove half flag, as I see not detection. I have gpu but half flag never worked for me even in your previous videos. Please help

[@vinayreddy8982]

Sir im getting proper output but in video the output classes is not starting from 1 can you please help it

[@brunospfc8511]

Thank man, at this point i do own you some money :D Great tutorials !

[@ammar46]

Hey, I have GPU, but half flag never works, so usually I keep it as false. This time the FPS is 0. Can I know how to solve it??

[@bottabhanuprakash]

we can track objects manually in this code?

[@nikzadmotamedi9393]

Hi Haroon, Could you please guide me how to connect a Hikrobot camera to the Yolo algorithm?

[@AryanDalal-qr9ru]

Hey Thanks for the video. Can you make the video while tracking with deep sort algo it is said to be have significant improvement comparatively. Thanks

[@JakubPeka1]

Ok, maybe a dumb question - but I would like to export all tracks and statistics so after for ex 30minutes of video I can see all tracks of people that were in my wideo, and export only tracks. is it possible? So i know how many people moved across my wideo, and I can see all the tracks so I can make a heatmap of those later on?

[@TheCodingBug]

Yes you can add a xode that wrotes id and tracks to a text file and then write a separate script that reads the text file and show visualization. Don't assign same ID to anyone again.

[@JakubPeka1]

@@TheCodingBug thanks ! I will give it a try!

[@oanminhquan1426]

I have a question: Is it possible to do live tracking like track directly from your camera ?

[@TheCodingBug]

Yes. Use --source 0 instead of video name.

[@apnabazar-w9m]

would you consider adding tag to your repo for skipping frames or vid_stride for better performance ie like process every 5 frame or every 10th frame for the 720p models for Realtime performance

[@zawarudo1818]

For some reason i get no detection. I tried with the pretrained model and my own model that was trained with custom dataset. Any ideas why?

[@pavantripathi1890]

Great video! Thanks

[@mosesdaudu]

Great Video. Pls how do i save the results in a json or txt file. i can see from the code that there is a "parser.add_argument('--save-txt')", however this doesn't save the results

[@gemshunt9637]

Hi! I have watched a lot of your videos and they are great! But in most cases it turns out "pth" format, how i can converted detectron2 r50-fpn model to "onnx"? I want to use this in opencv. Help me please!!!!!!!

[@AmitKumar-hm4gx]

Also, maybe a video (because I don't have fair idea about it and if you think that is something useful) on if we can use the generated traced.pt model every time there is an execution to load it via Torchscript to see if we can have faster inference. I have no idea if it does the same job as exporting the model to onxx-runtime and then maybe convert it to C++ to have faster inference.. maybe it does. Also, apologies if this is out of scope of the contents that you have thought for your channel.

[@rakeshkumarrout2629]

Hi this is really so useful.How to get the location after predicting output images?

[@TECHNEWSUNIVERSE]

did u find the solution?

[@rakeshkumarrout2629]

@@TECHNEWSUNIVERSE not yet

[@nithinhs7231]

Hi sir, I am getting this error how to resolve this? (yolov7_tracking) C:\Nithin_Data\workspace\yolov7>python detect_or_track.py --weights yolov7.pt --no-trace --view-img --nosave --source street.mp4 Traceback (most recent call last): File "C:\Nithin_Data\workspace\yolov7\detect_or_track.py", line 19, in from sort import * ModuleNotFoundError: No module named 'sort'

[@khaledmerit9417]

Have you course in udemy or corssera if yes the link please

[@TheCodingBug]

Hi. No I don't have any as of now.

[@acrtalksmalayalam8497]

Where can I run this yolov7 model other than computer

[@muh.fachrisyam6464]

Thanks for the video.. can you do tutorial to implement yolov7 into android camera?

[@rayyyy553]

hello thank u for the video ! do you know how i could obtain the coordinates (x,y) of the boxes ? i use yolov7 to detect traffic lights and i want to crop the traffic lights of the images (so crop the boxes that surrounded the traffic lights on the image) and give it to a CNN to know if it is red or green.

[@yusuftalhacam2242]

hello did you find out how to do this?

[@rayyyy553]

@@yusuftalhacam2242 yes ,wheb you open yolov7, you Can see in the desk "detect.py" ans you have in this programme xyz that ,i'm not sûre " correspond to thé coordinates.

[@tanweerbilal9077]

Hallo, The problem i am facing is that My person tracking id changes throughout the video sequence . What is the solution for that. Some details about video is that it is 3 minutes long video and when framed it has more than 5000 frames . for example in frame 1 Person A id =1 , and then after 50 frames he gets random id like 12 and throughout the frames it keeps changing.. Please if you have any solution for this problem and can you share it ? I have 5 persons in the video sequence and the id keep changes for every person . Thank you in advance

[@TheCodingBug]

Do these persons vanish from the camera for couple of seconds?

[@tanweerbilal9077]

@@TheCodingBug yes one person do vanish for some seconds and come back his Id also changes and some persons id changes while they are still in the frame. So the id keeps changing

[@mrtnlmao]

@@tanweerbilal9077 then it's obviously the IDs gonna change... also if in some frame the algorithm cant keep the classification, it detects the object, not specific persons... if you really want do make the unique IDs for each person, you gonna need model trained on your own dataset containing separated objects - in this case the faces of persons

[@Malinikripya]

How to change from SORT to DEEPSORT in the process

[@dankYbat]

Hello, how do you create new neural networks and wanted to give you an idea for creating new videos. Create a neural network for the game. For example, CS:GO. Howdy Ho did something similar on his KZbin channel, but he didn't show how it works. I would like to make a neural network for another shooter game, but I don't understand how all the files are connected. The neural network knows how to find the enemy, aim and shoot. He used mss library for this (to create screenshots from the game itself) and at this point I had problems (. I think this is a great idea and it will be very interesting to see such content from you. Thank you

[@AmitKumar-hm4gx]

Hello, your tracker code is based on RizwanMunawar's code of v7 and SORT - the moment I saw the trail lines for tracker, I knew it was from there because I already found that some weeks before. However, it's great to see that you are giving credits which people should. I don't know if you saw it but someone copied your v7 detection stuff and run it on some YT channel, I commented about it on one of your yolov7 videos. Secondly, can you please test running KZbin video - what happens when I do it, the video never ends. LoadStreams class (under utils->datasets) it uses threading concept when reading videos from stream, I believe it has to do something with this maybe the queue (separate thread) is unable to know when the video is over. When I run the YT video for detect and track -> I find myself left with black screen once the video is over and since the execution doesn't stop, the video doesn't get saved. The output is only saved if I interrupt manually (by pressing Q key).

[@TheCodingBug]

You are right. It is based on Rizwan Munawar's SORT code, along with my own modifications. He does not use class specific colors for tracking lines :). I made this code publicly available as well along with credits to the author/s (as you rightly noted). I have not seen the copied videos on some other KZbin channel. Maybe I missed your comment. I will check that out. Thank you for point out.

[@AmitKumar-hm4gx]

@@TheCodingBug yes the use of flags is a great addition has ton of flexibilities python detect_or_track.py --weights yolov7.pt --source kzbin.info/www/bejne/jafLiqh6q61kbas (you can adapt to your setting as per your flags) I believe you should get a blank/black screen as well and have the code not finish it's execution. If possible, do let me know if it worked out for you without any issues or if you were to resolve it.

[@at3411]

Hello Sir, In an earlier video you taught to run YOLOv7 on a custom dataset. Is it possible to track the custom classes?

[@majkel1381]

yes it is possible, just train your yolov7 model and use it with yolostream or sort

[@s4ifbn]

thank u

[@Studio-gs7ye]

-- no-trace why it is use in code? what it means at 3:21

[@lalithaee2378]

Can you add mask detection to this..

[@maxbunnies]

Awesome how can I track a feed on my monitor?

[@TheCodingBug]

Do you mean webcam feed?

[@maxbunnies]

@@TheCodingBug yes kinda. I mean say for example I have any type of video on the screen

[@michelangelo24]

Hi man, I loved the video. Could you someday make a tutorial teaching how to use Coqui TTS or Tacotron2 to clone voices in English and other languages? I gave up because I couldn't do it, and there are not many tutorials on how to use it. And the ones that do exist are not good. Please 😊

[@stianj6949]

Hi this worked, but have a problem.. The logging function "--save-txt" does not make a txt file only a folder called labes any way to fix this?

[@stianj6949]

Figured it out. Added this to detect_or_track.py under def detect #Write to file for *xyxy, conf, cls in reversed(det): if save_txt: # Write to file xywh = (xyxy2xywh(torch.tensor(xyxy).view(1, 4)) / gn).view(-1).tolist() # normalized xywh line = (cls, *xywh, conf) if opt.save_conf else (cls, *xywh) # label format with open(txt_path + '.txt', 'a') as f: f.write(('%g ' * len(line)).rstrip() % line + ' ')

[@vojtechcicha]

@@stianj6949 thanks man, that works. Do you know how can I add there also the ID of tracked object?

[@warrior_1309]

Great Video Sir , if possible , can you creat next video tracking using deepsort .

[@monikasingh5132]

I am unable to open yolov7.pt file

[@tanweerbilal9077]

Thanks for the video Can you please share the colab link . As i Dont have System with GPU and want to use your git on colab . Can you please share a colab guide please. Thanks

[@vishnum7985]

I am running this on my personal laptop CPU. My FPS is very low ~1 or 2. How can I increase it?

[@TheCodingBug]

Use GPU.

[@agusgistex1151]

my computer freezes while installing "pip install -r requirements_gpu.txt", is this normal?@@TheCodingBug Please Help 😭

[@TheCodingBug]

@@agusgistex1151 no this is not supposed to happen. Maybe just install requirements.txt and then install torch gpu from official website and skip installing requirements_gpu.txt

[@anime_on_data7594]

how to do yolov7 object detecion in ROI?

[@waynechang5571]

3:31 my prompt with no error, but output video doesn't have object detection box

[@ammar46]

Hey. Did you solve it??

[@ammar46]

It probably might be because of the half flag

[@ministerking2430]

谢谢

[@brunospfc8511]

Hey Bug, do yoloV8 ! Thanks

[@TheCodingBug]

On it.

[@FyrekittenIsCool]

gyat tracker

[@RandomGuy-df1oy]

Great video! Thanks

[@lalithaee2378]

Can you add mask detection to this.

[@pranavash3019]

i need the same