Cloned and used from The Official YOLOv7 Repository
Implementation of paper - YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors
To run the model, download the weights file from here and place in the root
folder.
Run these commands one by one when the Conda terminal is active:
#Create conda environment and install dependencies
conda create -n yolov7 python=3.9
conda activate yolov7
pip install -r requirements.txt
pip install -r requirement_gpu.txt # for gpu
python train.py --workers 1 --device 0 --batch-size 8 --epochs 100 --img 640 640 --data data/coco.yaml --hyp data/hyp.scratch.custom.yaml --cfg cfg/training/yolov7-w6.yaml --weights yolov7_training.pt
#increase batch size if gpu memory is high (>12GB)
Training will roughly take 6 hours or more depending on your GPU (6 hours on RTX 2060)
After training move the best.pt file from runs/train/exp/weights to the root folder where the repository resides
python test.py --weights best.pt --data data/coco.yaml --img 640 --iou 0.65 --device 0 --batch-size 8 --task test --save-txt --save-conf
#increase batch size if GPU memory is high
#Change IOU confidence level as per need, default is 0.65