Training custom YOLO v3 object detection model

In my previous tutorials, I showed you, how to simply use YOLO v3 object detection with the TensorFlow 2.x application and how to train Mnist custom object detection. At the end of the tutorial I promised, that I will show you how to train custom object detection. It was a challenging task, but I found a way to do that. However, before training a custom object detector, we must know where we may get a custom dataset or how we should label it, so in this tutorial, I will show you where you could get labeled dataset, how to prepare it for training, and finally how to train it!

✅ Support My Channel Through Patreon:

✅ One-Time Contribution Through PayPal:

[YOGESHPATIL-ptxn]

One of the best and easiest tutorials one could find over the internet.
There is only one thing in the code that I would like to correct.
Problem: One cannot train the model using the checkpoints saved.
Solution: Edit and add these lines in train.py on line number 58
TRAIN_FROM_CHECKPOINT_NAME =

i.e. replace the try block with the above solution in "if TRAIN_FROM_CHECKPOINT:" condition at line number 56

[ihavetallbrains]

Thank you so much for this awesome tutorial

[bilalbataineh]

Thank you so much for this great work.

[karimtamani]

This Guy right here deserve some credit, You are doing great Job my friend Thank you
But please i want to understand the code in details, do you have the documentation of the Tensorflow2.x Yolov3 Code

[rog]

great video !! you're doing a good job :D
also, can u make a tutorial on custom unet, basically how to prepare custom data for unet segmentation, it would be great if you help with this.!!

[nasimthander]

Why my precision and recall showing an array of values inside results.txt

[bushradayo]

Bro can you tell me how to plot PR curve after i have trained the model yolov3 on my own dataset? Need assistance in its programming, please help

[korlakuntasaikamal]

Hai . its very good tutorial . but i am not able to find out the custom weight file [number plate, car, truck, traffic signs....] will u pls send me the link of that custom weight file .

[lingyau]

thanks for great videos! however, i keep on getting error when downloading images with "py main.py downloader ... " i added environment variable already but still does not work. can you please help?
error: 'aws' is not recognized as an internal or external command,
operable program or batch file.

[taube-au]

Hello! Thanks for your video, it's very interesting and your description is clear, I like it.
Now I'm trying to repeat your tutorial and have some problem with it. It happens when I download image classes by terminal (If it's important, I changed command "python" for "python3" because of mac's installed python2). So, I want to know how my error depends from this solution. What version of Python do you use?
Thx for your answer and tutorials. ;)

[syauky]

i got error when run main.py from oidv4 toolkit. can you help me?

[accimtcomm]

Hello thanks for the lesson.. But It is not clear whether yolov3. weight is to be downloaded or it will be created by this code.

[aonguyenhung]

I don't see you putting custom class into coco.names file, is that neccessary ?

[syedfahad]

Can you please tell me what content is present inside tge file of Dataset_name.txt

[pranavk]

i want a code for helmet detection using yolov3 (both training & testing code), its Normal Yolov3 and NOT Yolo tiny or something, i do have training code but not sure if it has some errors, i also have a custom dataset labelled and ready, can anyone please help me with testing code and what things to add/edit in it since i am stuck and i have to submit the project in my college, its urgent.

Btw i am using Google Collab on windows 10 and code is based on Darknet53 not Tensorflow or Keras, so according to that can someone help please

[jaganisagar]

Hello Sir
Nice tutorial But I need your help for how to train dataset without GPU IN YOLOV3 and TF2 Please suggest me.
Please help me.
I run on aws server with micro instance without GPU. That time killed all process and not train one step.
Please give your suggestion.

[abcxyzxyz]

2020-09-04 15:14:16.410087: E failed call to cuInit: CUDA_ERROR_NO_DEVICE: no CUDA-capable device is detected

I'm getting this error.

[cbt]

Thank you for your awesome lessons. By the way, I don't think it is a good idea without using memory for training, it is so insanely slow that I can't even finish training in one week by measured time. But I don't have enough memory size either, unfortunately so I can't even use memory for image loading. So, as a last option, I want to use feature extractions to save into the disk to avoid I/O for loading data, do you have any idea how to do this?

[avinashgunda]

Hey, when Im running "python main.py downloader --classes and so on it is throwing an error like this
Traceback (most recent call last):
File "main.py", line 37, in <module> bounding_boxes_images(args, DEFAULT_OID_DIR)
File "W:\TensorFlow-2.x-YOLOv3-master\OIDv4_ToolKit\modules\bounding_boxes.py", line 56, in bounding_boxes_images class_code = df_classes.loc[df_classes[1] == class_name].values[0][0] IndexError: index 0 is out of bounds for axis 0 with size 0

I know that the size is empty or 0, can you suggest what to do at this point

[AhmedSayedAtwa]

Amazing Tutrial ..
Does anyone know the output file and "yolov3_custom.index" .. how can i use this output for other different applications? it works only with "detection_custom.py" .. can i convert it to yolov3.weights or " trained_weights_final.h5" so i can use for aimbot or another real-time detection applications that i already have ?