🔒 I need help updating a yolo model

[sullen berry]

I am currently working on a project which includes yolo object detection. I am currently using a database in yolov5, but I would like to change it to a different database in yolov8. I have the database, I am just unsure how to implement it into my project to replace the current one.

[unkempt skiff]

@sullen berry

Python help channel opened

Remember to:

• Ask your Python question, not if you can ask or if there's an expert who can help.
• Show a code sample as text (rather than a screenshot) and the error message, if you've got one.
• Explain what you expect to happen and what actually happens.

:warning: Do not pip install anything that isn't related to your question, especially if asked to over DMs.

Closes after a period of inactivity, or when you send !close.

[sudden kayak]

So, by database do you mean weights?

[sullen berry]

self.model = torch.hub.load('ultralytics/yolov5', 'custom', path='lib/best.pt', force_reload = True)

this is how the database is loaded at the moment

there is just a file named best.pt

no weights

but i would like to replace yolov5 with yolov8 and use a new pretrained database

which i have

are you able to help? @sudden kayak

[sudden kayak]

So the .pt file is usually regarded as the weights. It's not really a database as most people would think of a database.

You should just be able to repalce the path= argument with the new pre-trained weights file

[sullen berry]

i dont have the weights file

[sullen berry]

i need help transferring it to a weights file and implementing it into the code in replacement of this

[echo verge]

The .pt file is the weights file

[sudden kayak]

For yolov8, what files do you have with it?

[sullen berry]

https://universe.roboflow.com/james-pakis/fortnite-player-detection-hpiac/dataset/8

this is the dataset i would like to use

[echo verge]

Is this for malicious purposes?

[sullen berry]

no

[sudden kayak]

So you want to train and create your own new weights file using yolov8?

[sullen berry]

yes

[sudden kayak]

what's your goal with using this dataset in particular?

[sullen berry]

me and my friend are creating an ai detection for fortnite, its a project between only us two and will not be made public

there is no "malicious purpose" of this project, its just experimental

[winter falcon]

does it actually have any input on the game or is inserted in the game itself in anyway or is it being fed a screen capture?

[sullen berry]

there are no connections with the game whatsoever

it doesnt tamper with any files, it just detects players in a snippet of the screen

its very interesting to us

[winter falcon]

alright, i was asking because if it did it would break their tos making it so we arent allowed to help. as long as its only a snippet its fine

[sullen berry]

yeah there are no tampers with the game

we arent that stupid

[sudden kayak]

So, if you want to train a new thing with yolov8 and that dataset, then you need to replace the first argument with the actual yolov8 model. The starting weight file don't make a huuuge difference tbh if you're going to re-train if you tune your training well enough.

[sullen berry]

can you please help me with this? i have tried to replace the line of code with my own but it didnt work

[sudden kayak]

what did you replace it with?

[sullen berry]

i cant remember, i have deleted it

i can try retrace it

[sudden kayak]

Well, do you have the yolov8 model downloaded/accessible?

[sullen berry]

yes

[sullen berry]

here

i have downloaded it as a zip file

[sudden kayak]

So that's the weights file using the format yolov8 needs, but it's not the actual yolov8 model itself. Just the weights file trained with that dataset.

[sullen berry]

im sorry that i am so confused

how would i form the model

from the weights file

i need to make this into a model (.pt file)

https://universe.roboflow.com/james-pakis/fortnite-player-detection-hpiac/dataset/8

and integrate it so that pt file replaces the one in this

self.model = torch.hub.load('ultralytics/yolov5', 'custom', path='lib/best.pt', force_reload = True)

and change yolov5 to yolov8

[sudden kayak]

You can't. You need to download the yolov8 model from wherever it's originally hosted. The model can create the weights file after a sufficient amount of training.

From this: self.model = torch.hub.load('ultralytics/yolov5', 'custom', path='lib/best.pt', force_reload = True)

You should just need to repalce ultralytics/yolov5 with ultralytics/yolov8 (https://github.com/ultralytics/ultralytics)

[sullen berry]

i will try that now

doesnt work

[sudden kayak]

Ah, it looks like ultralytics dropped torch hub support for v8. You'll have to follow their docs to load in yolov8 https://docs.ultralytics.com/modes/train/#resuming-interrupted-trainings

[sullen berry]

so would i replace

self.model = torch.hub.load('ultralytics/yolov5', 'custom', path='lib/best.pt', force_reload = True)

with

model = YOLO('best.pt') # load a partially trained model

after i have imported yolo from ultralytics of course

[sudden kayak]

I'm not familiar with using their platform, you'd have to follow their guide for using it

[sullen berry]

ok thank you for the help

[unkempt skiff]

Python help channel closed

This help channel has been closed and it's no longer possible to send messages here. If your question wasn't answered, feel free to create a new post in #1035199133436354600. To maximize your chances of getting a response, check out this guide on asking good questions.