allen-engi/cv-yolov5

yolov5 complete code for cv team

Updated Version of YOLOv5 with CUDA Fix

Description

• Integrated from yolov5 official repo, fixed the issue on cuda core can't run on yolov7 model. Tested on conda environment with python3.9, pytorch=1.11.0, cudatoolkit=1.13

Installation on Windows

Conda environment -- Anaconda https://www.anaconda.com/
Python -- Python 3.9 installed with Anaconda

Install Anaconda

Select Installation Type : 'just me'
Anaconda Install Location : Anywhere you want, doesn't have to be on C drive
Advanved Installation Options :

Clone yolov5 Repo

Run git clone https://github.com/FlyerJB/YOLOv7-RoboMaster.git on your command prompt to some dir under C:/ drive or your OS drive to avoid Enviornment failure \

Make Conda Environment

Open Conda Command Prompt with Admin Right
Cd into yolov5 dir with cd <where you clone your yolov5>
And craete Conda Environment Conda Create -n <The Name You Like> Python3.9

Activate Conda Environment

Run command Conda activate <The Name You put from previous step>

pip install required packages

Option 1 : Install yolov5 for training on CPU
pip install -r requirements.txt

Option 2 : Install yolov5 for training on RTX GPU
pip install -r requirement_nv_gpu.txt
pip install torch==1.11.0+cu113 torchvision==0.12.0+cu113 torchaudio==0.11.0 --extra-index-url https://download.pytorch.org/whl/cu113

Validate Cuda Installation ( required for nv_gpu training )

Run python or python3 or py to run python
Run import torch
Run torch.cuda.is_available()
If it returns True, it means CUDA is successfully install on your device with Pytorch.
Then run exit() to exit python

Seperate Dataset

Validate set

10% - 20% of the whole dataset
It goes under data\val foler with train and label folders inside

Train set

80% to 90% of the whole dataset
It goes under data\train foler with train and label folders inside

Trained model

Our own model have been up uploaded to cv-yolov5/model files Models

Training

With GPU training

# train models
python train.py --workers 1 --device 0 --batch-size -1 --epochs 50 --img 640 --data data/coco_custom.yaml --hyp data/hyp.scratch.custom.yaml --cfg cfg/training/yolov5-custom.yaml --name yolov5-tut3 --weights yolov5.pt  

With CPU training

# train models
python train.py --workers 8 --device CPU --batch-size -1 --epochs 50 --img 640 --data data/coco_custom.yaml --hyp data/hyp.scratch.custom.yaml --cfg cfg/training/yolov5-custom.yaml --name yolov5-tut3 --weights yolov5.pt