Welcome to YOLOv3 Discussions!

[glenn-jocher]

👋 Welcome!

We’re using Discussions as a place to connect with other members of our community. We hope that you:

• Ask questions you’re wondering about.
• Share ideas.
• Engage with other community members.
• Welcome others and are open-minded. Remember that this is a community we
  build together 💪.

To get started, comment below with an introduction of yourself and tell us about what you do with this community.

[ShostNazar]

Hi. I used yolov3 with 1080TI . Now I have updated the video card to rtx 3080. And the program stuck .
Model Summary: 225 layers, 6.25733e+07 parameters, 6.25733e+07 gradients
i already deleted cuba and instaled it again. but didnt helped.

[ShostNazar]

my friend tried too on rtx 3080. and same didnt work.

[glenn-jocher]

@ShostNazar if your environment and CUDA drivers are set up correctly you should be able to train on all Turing and Ampere generation NVIDIA GPUs. I've trained successfully myself on earlier 1080Ti's, 2080Ti's and 3090 GPUs on the consumer card front, and a much wider range of enterprise GPUs on GCP and AWS. I'll post our environment issue reply below, which is based on YOLOv5 but also applies to this YOLOv3 repository as they now share the same codebase.

Environment Issues

It appears you may have environment problems. Please ensure you meet all dependency requirements if you are attempting to run YOLOv5 locally. If in doubt, create a new virtual Python 3.8 environment, clone the latest repo (code changes daily), and pip install -r requirements.txt again. We also highly recommend using one of our verified environments below.

Requirements

Python 3.8 or later with all requirements.txt dependencies installed, including torch>=1.7. To install run:

$ pip install -r requirements.txt

Environments

YOLOv5 may be run in any of the following up-to-date verified environments (with all dependencies including CUDA/CUDNN, Python and PyTorch preinstalled):

• Google Colab Notebook with free GPU:
• Kaggle Notebook with free GPU: https://www.kaggle.com/ultralytics/yolov5
• Google Cloud Deep Learning VM. See GCP Quickstart Guide
• Docker Image https://hub.docker.com/r/ultralytics/yolov5. See Docker Quickstart Guide

Status

If this badge is green, all YOLOv5 GitHub Actions Continuous Integration (CI) tests are passing. These tests evaluate proper operation of basic YOLOv5 functionality, including training (train.py), testing (test.py), inference (detect.py) and export (export.py) on MacOS, Windows, and Ubuntu.

[ShostNazar]

yes program see that i using 3080. but stuck for 10 min and do nothing. before all worked perfect

[glenn-jocher]

@ShostNazar it appears you are trying to run the archive branch. We do not offer support or maintain this branch anymore. I would recommend trying the master branch, which is now the default branch. The equivalent master branch command is:

git clone https://github.com/ultralytics/yolov3
cd yolov3
python detect.py --weights yolov3-spp.pt

Branch Notice

The ultralytics/yolov3 repository is now divided into two branches:

• Master branch: Forward-compatible with all YOLOv5 models and methods (recommended).

$ git clone https://github.com/ultralytics/yolov3  # master branch (default)

• Archive branch: Backwards-compatible with original darknet *.cfg models (⚠️ no longer maintained).

$ git clone -b archive https://github.com/ultralytics/yolov3  # archive branch

[daniramdani]

Hi. I my name's Dani, I used yolov3 for surveillance project, currently I test using a rtsp camera, I get 3 to 5 fps on processing image, to minimize resource usage I plan to decrease image processing to 1 image for every 5 second, is it able to configure using setting ? or need effort on coding side ?

[glenn-jocher]

@daniramdani you might try modifying the source here, though you may also want to implement a change detection intermediary or additional logic to suit your needs.

yolov3/utils/datasets.py

Lines 292 to 303 in 9f4e853

	def update(self, index, cap):
	# Read next stream frame in a daemon thread
	n = 0
	while cap.isOpened():
	n += 1
	# _, self.imgs[index] = cap.read()
	cap.grab()
	if n == 4: # read every 4th frame
	_, self.imgs[index] = cap.retrieve()
	n = 0
	time.sleep(0.01) # wait time

[glenn-jocher]

@daniramdani streamloader() is used for rtsp/http/rtmp streams and webcams only. loadimages() is used for all other media (images and videos).

[glenn-jocher]

@daniramdani the streaming dataloader is here:

yolov3/utils/datasets.py

Line 258 in 9f4e853

class LoadStreams: # multiple IP or RTSP cameras

Image and video dataloader is here:

yolov3/utils/datasets.py

Line 121 in 9f4e853

class LoadImages: # for inference

[daniramdani]

Oh ok, then. on checking. Thanks

[TheoGreenslade]

Hi. From what I've read about YOLO I thought that when calculating the loss an obj loss and noobj loss is calculated. I can't seem to find where this is done in the source code or where I would be able to scale the weighting on noobj loss up? Thank you for your help.

[glenn-jocher]

@TheoGreenslade in this implementation there is only objectness loss gain, and objectness loss positive sample gain (also applied only to positive samples). These values are in the hyperparameter file:

yolov3/data/hyp.scratch.yaml

Lines 16 to 17 in c1f8dd9

	obj: 1.0 # obj loss gain (scale with pixels)
	obj_pw: 1.0 # obj BCELoss positive_weight

[TheoGreenslade]

Ah great, thank you so much for your help!

[ShostNazar]

Do you have a version of yolo that was on your page on September 5? (I'm not a programmer and I wanted to just start train in one command like it was before). Because I have a lot of problems with the new version - i cant train on Windows 10. I get this error
Namespace(weights='yolov3.pt', cfg='', data='.\data\coco128.yaml', hyp='data/hyp.scratch.yaml', epochs=5, batch_size=16, img_size=[640, 640], rect=False, resume=False, nosave=False, notest=False, noautoanchor=False, evolve=False, bucket='', cache_images=False, image_weights=False, device='', multi_scale=False, single_cls=False, adam=False, sync_bn=False, local_rank=-1, log_imgs=16, log_artifacts=False, workers=8, project='runs/train', name='exp', exist_ok=False, quad=False, world_size=1, global_rank=-1, save_dir='runs\train\exp6', total_batch_size=16)
←[34m←[1mwandb: ←[0mInstall Weights & Biases for YOLOv3 logging with 'pip install wandb' (recommended)
Start Tensorboard with "tensorboard --logdir runs/train", view at http://localhost:6006/
←[34m←[1mhyperparameters: ←[0mlr0=0.01, lrf=0.2, momentum=0.937, weight_decay=0.0005, warmup_epochs=3.0, warmup_momentum=0.8, warmup_bias_lr=0.1, box=0.05, cls=0.5, cls_pw=1.0, obj=1.0, obj_pw=1.0, iou_t=0.2, anchor_t=4.0, fl_gamma=0.0, hsv_h=0.015, hsv_s=0.7, hsv_v=0.4, degrees=0.0, translate=0.1, scale=0.5, shear=0.0, perspective=0.0, flipud=0.0, fliplr=0.5, mosaic=1.0, mixup=0.0
fatal: not a git repository (or any of the parent directories): .git
←[34m←[1mgithub: ←[0mTraceback (most recent call last):
File "C:\Users\Nazar\Desktop\yolov3-master\train.py", line 519, in
train(hyp, opt, device, tb_writer, wandb)
File "C:\Users\Nazar\Desktop\yolov3-master\train.py", line 75, in train
attempt_download(weights) # download if not found locally
File "C:\Users\Nazar\Desktop\yolov3-master\utils\google_utils.py", line 30, in attempt_download
tag = subprocess.check_output('git tag', shell=True).decode().split()[-1]
File "C:\Users\Nazar\AppData\Local\Programs\Python\Python39\lib\subprocess.py", line 424, in check_output
return run(*popenargs, stdout=PIPE, timeout=timeout, check=True,
File "C:\Users\Nazar\AppData\Local\Programs\Python\Python39\lib\subprocess.py", line 528, in run
raise CalledProcessError(retcode, process.args,
subprocess.CalledProcessError: Command 'git tag' returned non-zero exit status 128.

[glenn-jocher]

@ShostNazar 👋 Hello, thank you for your interest in our work! This issue seems to lack the minimum requirements for a proper response, or is insufficiently detailed for us to help you. Please note that most technical problems are due to:

• Your modified or out-of-date code. If your issue is not reproducible in a new git clone version of this repo we can not debug it. Before going further run this code and verify your issue persists:

$ git clone https://github.com/ultralytics/yolov3 yolov3_new  # clone latest
$ cd yolov3_new
$ python detect.py  # verify detection

# CODE TO REPRODUCE YOUR ISSUE HERE

• Your custom data. If your issue is not reproducible in one of our 3 common datasets (COCO, COCO128, or VOC) we can not debug it. Visit our Custom Training Tutorial for guidelines on training your custom data. Examine train_batch0.jpg and test_batch0.jpg for a sanity check of your labels and images.

• Your environment. If your issue is not reproducible in one of the verified environments below we can not debug it. If you are running YOLOv5 locally, verify your environment meets all of the requirements.txt dependencies specified below. If in doubt, download Python 3.8.0 from https://www.python.org/, create a new venv, and install requirements.

If none of these apply to you, we suggest you close this issue and raise a new one using the 🐛 Bug Report template, providing screenshots and a minimum reproducible example of your issue. Thank you!

Requirements

Python 3.8 or later with all requirements.txt dependencies installed, including torch>=1.7. To install run:

$ pip install -r requirements.txt

Environments

YOLOv5 may be run in any of the following up-to-date verified environments (with all dependencies including CUDA/CUDNN, Python and PyTorch preinstalled):

• Google Colab and Kaggle notebooks with free GPU:
• Google Cloud Deep Learning VM. See GCP Quickstart Guide
• Amazon Deep Learning AMI. See AWS Quickstart Guide
• Docker Image. See Docker Quickstart Guide

Status

If this badge is green, all YOLOv5 GitHub Actions Continuous Integration (CI) tests are passing. These tests evaluate proper operation of basic YOLOv5 functionality, including training (train.py), testing (test.py), inference (detect.py) and export (export.py) on MacOS, Windows, and Ubuntu.

[TheoGreenslade]

Hi there!

I am slightly confused about how the models are generated from the yaml files. From my understanding, Darknet-53 has 53 convolutional layers however there are only 6 convolutional layers in the backbone part of the yaml file. Would you be able to clear this up for me. Additionally, in the comments in the yaml files, what do P1, P2, ..., P5 mean and the numbers next to them?

Thank you so much for your help.

[glenn-jocher]

@TheoGreenslade the YOLOv5 yamls are new type of configuration file, each line is a stage, and each stage can contain multiple layers.

P3, P4 etc. refers to the stride of the layer.

[beau-61]

Hello, I want to compare the performance of the model with and without Mosaic. How can I turn off the image preprocessing module of Mosaic?

[glenn-jocher]

@beau-61 mosaic probability is a hyperparameter that you can vary from 0 to 1 here:

yolov3/data/hyp.scratch.yaml

Line 32 in 47ac683

mosaic: 1.0 # image mosaic (probability)

[beau-61]

Hello, where can I modify the activation function, thank you very much

[beau-61]

from utils.activations import FReLU
...
self.act = FReLU() if act is True else (act if isinstance(act, nn.Module) else nn.Identity())
Forgive me, I modified it to look like the above, but the following error occurred：
Traceback (most recent call last):
File "C:/yolov5-master/train.py", line 541, in
train(hyp, opt, device, tb_writer)
File "C:/yolov5-master/train.py", line 88, in train
model = Model(opt.cfg or ckpt['model'].yaml, ch=3, nc=nc, anchors=hyp.get('anchors')).to(device) # create
File "C:\yolov5-master\models\yolo.py", line 91, in init
self.model, self.save = parse_model(deepcopy(self.yaml), ch=[ch]) # model, savelist
File "C:\yolov5-master\models\yolo.py", line 269, in parse_model
m_ = nn.Sequential(*[m(*args) for _ in range(n)]) if n > 1 else m(*args) # module
File "C:\yolov5-master\models\common.py", line 168, in init
self.conv = Conv(c1 * 4, c2, k, s, p, g, act)
File "C:\yolov5-master\models\common.py", line 39, in init
self.act = FReLU() if act is True else (act if isinstance(act, nn.Module) else nn.Identity())
TypeError: init() missing 1 required positional argument: 'c1'

[glenn-jocher]

@beau-61 FReLU() is missing 1 required positional argument: 'c1'

[beau-61]

It is，could you please help and tell me what should i do next？

[beau-61]

self.act = FReLU(c1=c1) if act is True else (act if isinstance(act, nn.Module) else nn.Identity())
I modified it to look like the above, but the following error occurred：
RuntimeError: Given groups=12, weight of size [12, 1, 3, 3], expected input[1, 32, 128, 128] to have 12 channels, but got 32 channels instead

It seems that the size of the last layer does not match, but my ability is not enough to modify the activation function of the layer alone.

[glenn-jocher]

@beau-61 c1 into your activation would be c2 coming out of the previous convolution.

[eaadithya]

Hello @glenn-jocher, I am trying use YOLOv3 with a classification task, could you please tell me where to find, how image sizes are handled. I could only find out that it checks if the image size is a multiple of stride. All my input images are 1024 square. According to what I read on YOLOv3 images has to be 416x416. But I've already got good results. Are all images later getting reduced to 416? Please help me with this. Thanks.

[glenn-jocher]

@eaadithya see train.py argparser for default settings:

yolov3/train.py

Lines 452 to 489 in 1be3170

	if __name__ == '__main__':
	parser = argparse.ArgumentParser()
	parser.add_argument('--weights', type=str, default='yolov3.pt', help='initial weights path')
	parser.add_argument('--cfg', type=str, default='', help='model.yaml path')
	parser.add_argument('--data', type=str, default='data/coco128.yaml', help='data.yaml path')
	parser.add_argument('--hyp', type=str, default='data/hyp.scratch.yaml', help='hyperparameters path')
	parser.add_argument('--epochs', type=int, default=300)
	parser.add_argument('--batch-size', type=int, default=16, help='total batch size for all GPUs')
	parser.add_argument('--img-size', nargs='+', type=int, default=[640, 640], help='[train, test] image sizes')
	parser.add_argument('--rect', action='store_true', help='rectangular training')
	parser.add_argument('--resume', nargs='?', const=True, default=False, help='resume most recent training')
	parser.add_argument('--nosave', action='store_true', help='only save final checkpoint')
	parser.add_argument('--notest', action='store_true', help='only test final epoch')
	parser.add_argument('--noautoanchor', action='store_true', help='disable autoanchor check')
	parser.add_argument('--evolve', action='store_true', help='evolve hyperparameters')
	parser.add_argument('--bucket', type=str, default='', help='gsutil bucket')
	parser.add_argument('--cache-images', action='store_true', help='cache images for faster training')
	parser.add_argument('--image-weights', action='store_true', help='use weighted image selection for training')
	parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
	parser.add_argument('--multi-scale', action='store_true', help='vary img-size +/- 50%%')
	parser.add_argument('--single-cls', action='store_true', help='train multi-class data as single-class')
	parser.add_argument('--adam', action='store_true', help='use torch.optim.Adam() optimizer')
	parser.add_argument('--sync-bn', action='store_true', help='use SyncBatchNorm, only available in DDP mode')
	parser.add_argument('--local_rank', type=int, default=-1, help='DDP parameter, do not modify')
	parser.add_argument('--workers', type=int, default=8, help='maximum number of dataloader workers')
	parser.add_argument('--project', default='runs/train', help='save to project/name')
	parser.add_argument('--entity', default=None, help='W&B entity')
	parser.add_argument('--name', default='exp', help='save to project/name')
	parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment')
	parser.add_argument('--quad', action='store_true', help='quad dataloader')
	parser.add_argument('--linear-lr', action='store_true', help='linear LR')
	parser.add_argument('--label-smoothing', type=float, default=0.0, help='Label smoothing epsilon')
	parser.add_argument('--upload_dataset', action='store_true', help='Upload dataset as W&B artifact table')
	parser.add_argument('--bbox_interval', type=int, default=-1, help='Set bounding-box image logging interval for W&B')
	parser.add_argument('--save_period', type=int, default=-1, help='Log model after every "save_period" epoch')
	parser.add_argument('--artifact_alias', type=str, default="latest", help='version of dataset artifact to be used')
	opt = parser.parse_args()

[eaadithya]

@glenn-jocher Thanks for your prompt reply. But I already added the argument '--img-size 1024' while training. I was trying to find out how this parameter affect the convolution and the number of outputs on the presumption that output size depends on input. Could you please help me figure out how it works with size 1024 input size.

[glenn-jocher]

@eaadithya YOLO output sizes are functions of the input size divided by output stride.