Train a custom object detector using YOLOv4 and YOLOv4-tiny (on Linux system)

This video shows step by step tutorial on how to train a custom YOLOv4 object detector using darknet on your local Linux machine. In this tutorial, I have trained a custom YOLOv4 and YOLOv4-tiny detector for mask detection.

① ⚡⚡ My Website Blog post on this ⚡⚡

⚡⚡ Medium post on this ⚡⚡

② ⚡⚡ My GitHub link for Linux custom YOLOv4 training files ⚡⚡

③ ⚡⚡ My GitHub link for Linux custom YOLOv4-tiny training files ⚡⚡

④ ⚡⚡ Github link for Official Darknet repository ⚡⚡

⑥ ⚡⚡ My custom trained yolov4 weights files for mask detection ⚡⚡

⑦ ⚡⚡ Pre-trained yolov4 weights file ⚡⚡

⑧ ⚡⚡ Pre-trained yolov4-tiny weights file ⚡⚡

#yolov4 #objectdetection #yolov4onlinux

▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬

▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
Making these videos takes a lot of time and effort, so if you like these videos and if you can, then please support the channel using any of the following:

► Buy me a coffee! ☕

► Support channel on Patreon! 🖖


▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
Thanks for watching!

[qq_teacher]

Thank you, it was very helpful for me!!

[samsical]

Thank you for the video, my friend. This is a great resource for learning!

[erenjaeger]

I trained my own custom dataset on yolov4 after referring your article on medium. Thanks for making that💯

[annapoorani]

Hi how to find the perfomance pf yolo i.e specificity sensitivity accuracy and all?

[MohitGupta-bngj]

Hey! Thanks for the great video!! I have a question. I know you might be busy but please reply when you get time. Can we train YOLOv4 from scratch?
What should be the command for training from scratch?
Thanks :)

[abubakarsaleem]

Thanks sir your content is amazing thats what i was searching for very a long time on internet, sir kindly its a request make the training videos and pretrained model videos on pytorch or tensorflow on ubuntu not on google colab.🙏

[hemanthkumarar]

Sir, amazing content. Thank you.
I am currently training my model which has been going for 10 hours and will continue for another 5 hours.
Sir, how can we detect multiple images inside a file together as a batch and output to a folder.
Instead of detecting each image every time, what should I do so that I can detect images in a batch after training?

Sir, I think the ideal solution would be to make a python script which does these single steps together in the background. Please share what do you think sir? Is it the best approach?

[merondesta]

that was very wonderful! how can i use tensor board with this then. i have already trained the model

[samuelmoronta]

Im following all steps of your tutorial but accidentally put less 4 classes, and it's meant to have 6 is in (obj.data file). So After finishing training process my yolov4.weight goes wrong in predictions results and drawing boxes too. I have a dataset of 12, 000 images. Do you think that little mistake causes wrong results?

[ragavanrs]

hello
I trained in ball detection using yolov4 I followed the same steps which you are taught in this video, but I am using nvidia dgx station, while testing/predicting the image worked fine but while testing the video I got an error.(cannot load image ../test_videos/football.mp4)

[RajSharma-uivq]

Great Video. BTW, do you know upto how many classes can be used in YOLOv4-tiny?

[nylcnylc]

I trained my own custom dataset (only 1 class: top down view of person) on yolov4 (Dataset cut from a video, about 400 images from same video), and my custom model can detect all persons on the tranining video correctly, but if i use the custom model to detect an image without the person, it will still detect many object as person with more than 0.5 confidence, is there any thing i did work?? Thanks

Another question is that, after training from the orignal data set. If i want to add more images to train, do i need to train from begining, or can train from the last training point?

Another question, the yoloV4 model have the person detection already, but top down view of human detection is not good, so i want to make it more accurate on top down view of human, then what should i do?

Thanks, i am a begininer. Whole you can help. thanks

[ragavanrs]

hello
I have trained a anpr model with less number of data sets (3500) on Nvidia dgx station (linux) with help of this video but whenever i trained with above 10 k data sets I got " (segmentation fault (core dumped)" this error . i have to train anpr model with 1 lakh data sets . how to train a large number of data on nividia dgx (linux)? pls help to slove this problem

[abubakarsaleem]

sir ( 10:08 minutes of video) you took max_batches = 6000 but why sir?? if we have 2 classes then if we multiply with (2classess*2000) it will become 4000, then why we put 6000 in max_batches.Kindly tell me sir.

[vadivelutrollcenter]

😓i am the unlucky person error occurs
I resolved core dump😅
But new error cuda error:out of memory😶

[randomshorts..]

Help me couldn't open file error😢