Learning Experience Sharing Seven: YOLOv5 Code Chinese Comments

There are many friends who are learning YOLOv5 recently. In order to make it easier for everyone to better understand the source code, YOLOv5 is annotated in Chinese.Friends who need it can follow my WeChat official account:Artificial Intelligence AIAlgorithm Engineer after private message.

The following is a partial code example.

def main(opt):# 1, logging and wandb initialization# log initializationset_logging(RANK)if RANK in [-1, 0]:# can output all training opt parametersprint(colorstr('train: ') + ', '.join(f'{k}={v}' for k, v in vars(opt).items()))# This code is used to check if the code version is the latestcheck_git_status()# Used to check whether the required packages in requirements.txt are all satisfiedcheck_requirements(exclude=['thop'])# wandb logging initializationwandb_run = check_wandb_resume(opt)# 2. Determine whether to use the breakpoint to resume the training and load the parametersif opt.resume and not wandb_run:# Use breakpoints to continue training to read relevant parameters from last.pt# If resume is str, it will indicate that the path address of the model is passed in# If resume is True, use the get_lastest_run() function to find the latest weight file last.pt in the folder where runs areckpt = opt.resume if isinstance(opt.resume, str) else get_latest_run()assert os.path.isfile(ckpt), 'ERROR: --resume checkpoint does not exist' # check# The relevant opt ​​parameters should also be replaced with the opt parameters in last.ptwith open(Path(ckpt).parent.parent / 'opt.yaml') as f:opt = argparse.Namespace(**yaml.safe_load(f)) # replaceopt.cfg, opt.weights, opt.resume = '', ckpt, True # reinstatelogger.info('Resuming training from %s' % ckpt) # printelse:# You can read relevant parameters from the file without using breakpoints to continue training# opt.hyp = opt.hyp or ('hyp.finetune.yaml' if opt.weights else 'hyp.scratch.yaml')opt.data, opt.cfg, opt.hyp = check_file(opt.data), check_file(opt.cfg), check_file(opt.hyp) # check files

If you think it is helpful to everyone, please like, collect, and follow. I will continue to provide reference for everyone to do experiments.Any questions are welcome to private message me.

If you need more program information and answer questions, please pay attention - WeChat public account: Artificial intelligence AIAlgorithm Engineer