YOLOv3 in PyTorch > ONNX > CoreML > TFLite

This guide explains how to train your own custom dataset with YOLOv3.

Before You Start

Clone this repo, download COCO dataset, and install requirements.txt dependencies, including Python>=3.7 and PyTorch>=1.4.

git clone https://github.com/ultralytics/yolov3
bash yolov3/data/get_coco2017.sh  # 19GB
cd yolov3
pip install -U -r requirements.txt

Train On Custom Data

1. Label your data in Darknet format. After using a tool like Labelbox to label your images, you'll need to export your data to darknet format. Your data should follow the example created by get_coco2017.sh, with images and labels in separate parallel folders, and one label file per image (if no objects in image, no label file is required). The label file specifications are:

• One row per object
• Each row is class x_center y_center width height format.
• Box coordinates must be in normalized xywh format (from 0 - 1). If your boxes are in pixels, divide x_center and width by image width, and y_center and height by image height.
• Class numbers are zero-indexed (start from 0).

Each image's label file must be locatable by simply replacing /images/*.jpg with /labels/*.txt in its pathname. An example image and label pair would be:

../coco/images/train2017/000000109622.jpg  # image
../coco/labels/train2017/000000109622.txt  # label

An example label file with 5 persons (all class 0):

2. Create train and test *.txt files. Here we create data/coco16.txt, which contains the first 16 images of the COCO2017 dataset. We will use this small dataset for both training and testing. Each row contains a path to an image, and remember one label must also exist in a corresponding /labels folder for each image containing objects.

3. Create new *.names file listing the class names in our dataset. Here we use the existing data/coco.names file. Classes are zero indexed, so person is class 0, bicycle is class 1, etc.

4. Create new *.data file with your class count (COCO has 80 classes), paths to train and validation datasets (we use the same images twice here, but in practice you'll want to validate your results on a separate set of images), and with the path to your *.names file. Save as data/coco16.data.

5. Update yolov3-spp.cfg (optional). By default each YOLO layer has 255 outputs: 85 values per anchor [4 box coordinates + 1 object confidence + 80 class confidences], times 3 anchors. Update the settings to filters=[5 + n] * 3 and classes=n, where n is your class count. This modification should be made in all 3 YOLO layers.

6. (OPTIONAL) Update hyperparameters such as LR, LR scheduler, optimizer, augmentation settings, multi_scale settings, etc in train.py for your particular task. If in doubt about these settings, we recommend you start with all-default settings before changing anything.

7. Train. Run python3 train.py --cfg yolov3-spp.cfg --data data/coco16.data --nosave to train using your custom *.data and *.cfg. By default pretrained --weights yolov3-spp-ultralytics.pt is used to initialize your model. You can instead train from scratch with --weights '', or from any other weights or backbone of your choice, as long as it corresponds to your *.cfg.

Visualize Results

Run from utils import utils; utils.plot_results() to see your training losses and performance metrics vs epoch. If you don't see acceptable performance, try hyperparameter tuning and re-training. Multiple results.txt files are overlaid automatically to compare performance.

Here we see training results from data/coco64.data starting from scratch, a darknet53 backbone, and our yolov3-spp-ultralytics.pt pretrained weights.

Run inference with your trained model by copying an image to data/samples folder and running
python3 detect.py --weights weights/last.pt

Reproduce Our Results

To reproduce this tutorial, simply run the following code. This trains all the various tutorials, saves each results*.txt file separately, and plots them together as results.png. It all takes less than 30 minutes on a 2080Ti.

git clone https://github.com/ultralytics/yolov3
python3 -c "from yolov3.utils.google_utils import gdrive_download; gdrive_download('1h0Id-7GUyuAmyc9Pwo2c3IZ17uExPvOA','coco2017demos.zip')"  # datasets (20 Mb)
cd yolov3
python3 train.py --data coco64.data --batch 16 --epochs 300 --nosave --cache --weights '' --name from_scratch
python3 train.py --data coco64.data --batch 16 --epochs 300 --nosave --cache --weights yolov3-spp-ultralytics.pt --name from_yolov3-spp-ultralytics
python3 train.py --data coco64.data --batch 16 --epochs 300 --nosave --cache --weights darknet53.conv.74 --name from_darknet53.conv.74
python3 train.py --data coco1.data --batch 1 --epochs 300 --nosave --cache --weights darknet53.conv.74 --name 1img
python3 train.py --data coco1cls.data --batch 16 --epochs 300 --nosave --cache --weights darknet53.conv.74 --cfg yolov3-spp-1cls.cfg --name 1cls

Reproduce Our Environment

To access an up-to-date working environment (with all dependencies including CUDA/CUDNN, Python and PyTorch preinstalled), consider a:

• GCP Deep Learning VM with $300 free credit offer: See our GCP Quickstart Guide
• Google Colab Notebook with 12 hours of free GPU time: Google Colab Notebook
• Docker Image from https://hub.docker.com/r/ultralytics/yolov3. See Docker Quickstart Guide

Hi, Thanks for sharing your work ! I would like what is your configuration for the training of yolov3.cfg to get 55% MAP ? We tried 100 epochs but we got a MAP (35%) who don't really change much more. And the test loss start diverge a little. Why you give a very high loss gain for the confidence loss ? Thanks in advance for your reply.

Does this repo. support CSPResNeXt50-PANet-SPP? (https://github.com/WongKinYiu/CrossStagePartialNetworks/)

AlexeyABs support: https://github.com/AlexeyAB/darknet/issues/4406

My tests have found it to be a clear winner over yolov3-spp in terms of mAP and speed.

Training hyperparameters in this repo are defined in train.py, including augmentation settings: https://github.com/ultralytics/yolov3/blob/df4f25e610bc31af3ba458dce4e569bb49174745/train.py#L35-L54

We began with darknet defaults before evolving the values using the result of our hyp evolution code:

python3 train.py --data data/coco.data --weights '' --img-size 320 --epochs 1 --batch-size 64 -- accumulate 1 --evolve

The process is simple: for each new generation, the prior generation with the highest fitness (out of all previous generations) is selected for mutation. All parameters are mutated simultaneously based on a normal distribution with about 20% 1-sigma: https://github.com/ultralytics/yolov3/blob/df4f25e610bc31af3ba458dce4e569bb49174745/train.py#L390-L396

Fitness is defined as a weighted mAP and F1 combination at the end of epoch 0, under the assumption that better epoch 0 results correlate to better final results, which may or may not be true. https://github.com/ultralytics/yolov3/blob/bd924576048af29de0a48d4bb55bbe24e09537a6/utils/utils.py#L605-L608

An example snapshot of the results are here. Fitness is on the y axis (higher is better). from utils.utils import *; plot_evolution_results(hyp)

This guide explains how to train your data with YOLOv3 using Transfer Learning. Transfer learning can be a useful way to quickly retrain YOLOv3 on new data without needing to retrain the entire network. We accomplish this by starting from the official YOLOv3 weights, and setting each layer's .requires_grad field to false that we do not want to calculate gradients for and optimize.

Before You Start

1. Update (Python >= 3.7, PyTorch >= 1.3, etc.) and install requirements.txt dependencies.
2. Clone repo: git clone https://github.com/ultralytics/yolov3
3. Download COCO: bash yolov3/data/get_coco2017.sh

Transfer Learning

1. Download pretrained weights from our Google Drive folder that you want to use to transfer learn, and place them in yolov3/weights/.

2. Update *.cfg file (optional). Each YOLO layer has 255 outputs: 85 outputs per anchor [4 box coordinates + 1 object confidence + 80 class confidences], times 3 anchors. If you use fewer classes, reduce filters to filters=[4 + 1 + n] * 3, where n is your class count. This modification should be made to the layer preceding each of the 3 YOLO layers. Also modify classes=80 to classes=n in each YOLO layer, where n is your class count.

3. Train.

python3 train.py --data coco1cls.data --cfg yolov3-spp-1cls.cfg --weights weights/yolov3-spp.pt --transfer

Run the above code to transfer learn on COCO, or specify your own data as --data data/custom.data (See https://github.com/ultralytics/yolov3/wiki/Train-Custom-Data).

If you created a custom *.cfg file, specify it as --cfg custom.cfg.

You can observe in the Model Summary (using model_info(model, report='full') in train.py) that only the 3 YOLO layers have their gradients activated now (all other layers are frozen for duration of training):

Reproduce Our Environment

To access an up-to-date working environment (with all dependencies including CUDA/CUDNN, Python and PyTorch preinstalled), consider a:

• GCP Deep Learning VM with $300 free credit offer: See our GCP Quickstart Guide
• Google Colab Notebook with 12 hours of free GPU time: Google Colab Notebook
• Docker Image from https://hub.docker.com/r/ultralytics/yolov3. See Docker Quickstart Guide

This guide explains how to train your own single-class dataset with YOLOv3.

Before You Start

1. Update (Python >= 3.7, PyTorch >= 1.3, etc.) and install requirements.txt dependencies.
2. Clone repo: git clone https://github.com/ultralytics/yolov3
3. Download COCO: bash yolov3/data/get_coco2017.sh

Train On Custom Data

1. Label your data in Darknet format. After using a tool like Labelbox to label your images, you'll need to export your data to darknet format. Your data should follow the example created by get_coco2017.sh, with images and labels in separate parallel folders, and one label file per image (if no objects in image, no label file is required). The label file specifications are:

• One row per object
• Each row is class x_center y_center width height format.
• Box coordinates must be in normalized xywh format (from 0 - 1). If your boxes are in pixels, divide x_center and width by image width, and y_center and height by image height.
• Class numbers are zero-indexed (start from 0).

Each image's label file must be locatable by simply replacing /images/*.jpg with /labels/*.txt in its pathname. An example image and label pair would be:

../coco/images/train2017/000000109622.jpg  # image
../coco/labels/train2017/000000109622.txt  # label

An example label file with 4 persons (all class 0):

2. Create train and test *.txt files. Here we create data/coco_1cls.txt, which contains 5 images with only persons from the coco 2014 trainval dataset. We will use this small dataset for both training and testing. Each row contains a path to an image, and remember one label must also exist in a corresponding /labels folder for each image that has targets.

3. Create new *.names file listing all of the names for the classes in our dataset. Here we use the existing data/coco.names file. Classes are zero indexed, so person is class 0.

4. Update data/coco.data lines 2 and 3 to point to our new text file for training and validation (in your own data you would likely want to use separate train and test sets). Also update line 1 to our new class count, if not 80, and lastly update line 4 to point to our new *.names file, if you created one. Save the modified file as data/coco_1cls.data.

5. Update *.cfg file (optional). Each YOLO layer has 255 outputs: 85 outputs per anchor [4 box coordinates + 1 object confidence + 80 class confidences], times 3 anchors. If you use fewer classes, reduce filters to filters=[4 + 1 + n] * 3, where n is your class count. This modification should be made to the layer preceding each of the 3 YOLO layers. Also modify classes=80 to classes=n in each YOLO layer, where n is your class count (for single class training, n=1).

6. (OPTIONAL) Update hyperparameters such as LR, LR scheduler, optimizer, augmentation settings, multi_scale settings, etc in train.py for your particular task. We recommend you start with all-default settings first updating anything.

7. Train. Run python3 train.py --data data/coco_1cls.data to train using your custom data. If you created a custom *.cfg file as well, specify it using --cfg cfg/my_new_file.cfg.

Visualize Results

Run from utils import utils; utils.plot_results() to see your training losses and performance metrics vs epoch. If you don't see acceptable performance, try hyperparameter tuning and re-training. Multiple results.txt files are overlaid automatically to compare performance.

Here we see results from training on coco_1cls.data using the default yolov3-spp.cfg and also a single-class yolov3-spp-1cls.cfg, available in the data/ and cfg/ folders.

Evaluate your trained model: copy COCO_val2014_000000001464.jpg to data/samples folder and run python3 detect.py --weights weights/last.pt

Reproduce Our Results

To reproduce this tutorial, simply run the following code. This trains all the various tutorials, saves each results*.txt file separately, and plots them together as results.png. It all takes less than 30 minutes on a 2080Ti.

git clone https://github.com/ultralytics/yolov3
python3 -c "from yolov3.utils.google_utils import gdrive_download; gdrive_download('1h0Id-7GUyuAmyc9Pwo2c3IZ17uExPvOA','coco2017demos.zip')"  # datasets (20 Mb)
cd yolov3
python3 train.py --data coco64.data --batch 16 --accum 1 --epochs 300 --nosave --cache --weights '' --name from_scratch
python3 train.py --data coco64.data --batch 16 --accum 1 --epochs 300 --nosave --cache --weights yolov3-spp-ultralytics.pt --name from_yolov3-spp-ultralytics
python3 train.py --data coco64.data --batch 16 --accum 1 --epochs 300 --nosave --cache --weights darknet53.conv.74 --name from_darknet53.conv.74
python3 train.py --data coco1.data --batch 1 --accum 1 --epochs 300 --nosave --cache --weights darknet53.conv.74 --name 1img
python3 train.py --data coco1cls.data --batch 16 --accum 1 --epochs 300 --nosave --cache --weights darknet53.conv.74 --cfg yolov3-spp-1cls.cfg --name 1cls

Reproduce Our Environment

To access an up-to-date working environment (with all dependencies including CUDA/CUDNN, Python and PyTorch preinstalled), consider a:

• GCP Deep Learning VM with $300 free credit offer: See our GCP Quickstart Guide
• Google Colab Notebook with 12 hours of free GPU time: Google Colab Notebook
• Docker Image from https://hub.docker.com/r/ultralytics/yolov3. See Docker Quickstart Guide

Created custom files for training: (4 + 1 + 13) * 3 = 54 13 classes, 54 filters

*.names has 13 names in it *.cfg was converted properly w 13 for classes, 54 for filters in all 3 yolo blocks

yolov3/utils/utils.py", line 451, in build_targets assert c.max() <= model.nc, 'Target classes exceed model classes' AssertionError: Target classes exceed model classes

Hey I get a new error whan I run the train script:

Downloading https://drive.google.com/uc?export=download&id=158g62Vs14E3aj7oPVPuEnNZMKFNgGyNq as weights/ultralytics49.pt... Done (2.8s)
Traceback (most recent call last):
  File "train.py", line 444, in <module>
    train()  # train normally
  File "train.py", line 111, in train
    chkpt['model'] = {k: v for k, v in chkpt['model'].items() if model.state_dict()[k].numel() == v.numel()}
  File "train.py", line 111, in <dictcomp>
    chkpt['model'] = {k: v for k, v in chkpt['model'].items() if model.state_dict()[k].numel() == v.numel()}
KeyError: 'module_list.85.Conv2d.weight'

The original darknet learning rate (LR) scheduler parameters are set in a model's *.cfg file:

• learning_rate: initial LR
• burn_in: number of batches to ramp LR from 0 to learning_rate in epoch 0
• max_batches: the number of batches to train the model to
• policy: type of LR scheduler
• steps: batch numbers at which LR is reduced
• scales: LR multiple applied at steps (gamma in PyTorch)

In this repo LR scheduling is set in train.py. We set the initial and final LRs as hyperparameters hyp['lr0'] and hyp['lrf'], where the final LR = lr0 * (10 ** lrf) . For example, if the initial LR is 0.001 and the final LR is 100 times (1e-2) smaller, hyp['lrf']=0.001 and hyp['lrf']=-2. This plot shows two of the available PyTorch LR schedulers, with the MultiStepLR scheduler following the original darknet implementation (at batch_size=64 on COCO). To learn more please visit: https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate

The LR hyperparameters are tunable, along with all the rest of the model hyperparmeters in train.py:

https://github.com/ultralytics/yolov3/blob/1771ffb1cf293ef176ac6ef2ddb7af7ca672c0e7/train.py#L13-L25

Actual LR scheduling is set further down in train.py, and has been tuned for COCO training. You may want to set your own scheduler according to your specific custom dataset and training requirements, and also adjust it's hyperparameters accordingly. https://github.com/ultralytics/yolov3/blob/bd2378fad1578e7d7722ad846458ad7a2bb43442/train.py#L102-L109

Thanks for your improvement of this YOLOv3 implementation. I have just test the training ,got some problem . I follow these steps.

1. load the original yolov3.weight to the model
2. train it on coco2014 with your train.py. 3.Got the following logs ,the precision is down fast from 0.5->0.1. but recall is up to 0.35. see Screenshot here

4.I save the weight with precision0.2, and run the detect.py the result like this , if I do not train,the orginal wight can get this result:

I do not know whether I used wrong parameters or something else, lead to generation of many bbox . could you give me some suggestion? Thank you~

Dear all,

Apparently it is not possible to train the model with gray images.

Even if I convert the images in the getitem function, then the code will explode due to the multiple dependencies on the channel. E.g

    bs, _, h, w = imgs.shape  # batch size, _, height, width

Cheers,

Francesco Saverio

Bumps cirrus-actions/rebase from 1.7 to 1.8.

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.

Search before asking

• [X] I have searched the YOLOv3 issues and found no similar bug report.

YOLOv3 Component

No response

Bug

The docker image appears to not have a functioning python environment.

Environment

ultralytics/yolov3 docker container

Minimal Reproducible Example

I start by running the image in interactive mode: docker run --ipc=host -it --gpus all ultralytics/yolov3:latest Now run the training script: python train.py gives the following output:

Traceback (most recent call last):
  File "train.py", line 34, in <module>
    import val  # for end-of-epoch mAP
  File "/usr/src/app/val.py", line 26, in <module>
    from models.common import DetectMultiBackend
  File "/usr/src/app/models/common.py", line 13, in <module>
    import cv2
  File "/opt/conda/lib/python3.8/site-packages/cv2/__init__.py", line 181, in <module>
    bootstrap()
  File "/opt/conda/lib/python3.8/site-packages/cv2/__init__.py", line 175, in bootstrap
    if __load_extra_py_code_for_module("cv2", submodule, DEBUG):
  File "/opt/conda/lib/python3.8/site-packages/cv2/__init__.py", line 28, in __load_extra_py_code_for_module
    py_module = importlib.import_module(module_name)
  File "/opt/conda/lib/python3.8/importlib/__init__.py", line 127, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
  File "/opt/conda/lib/python3.8/site-packages/cv2/mat_wrapper/__init__.py", line 33, in <module>
    cv._registerMatType(Mat)
AttributeError: partially initialized module 'cv2' has no attribute '_registerMatType' (most likely due to a circular import)

Additional

No response

Are you willing to submit a PR?

• [ ] Yes I'd like to help by submitting a PR!

Search before asking

• [X] I have searched the YOLOv3 issues and discussions and found no similar questions.

Question

waiting for your kindness help!!!

I load pretrained yolov3 model from torch hub, like below:

model = torch.hub.load('ultralytics/yolov3', 'yolov3')

And I trained the model on my custom data, I really got the best.pt and last.pt weights file in the folder of runs/train/exp60/weights. However, when I try to load the model use my own weights files, there is something wrong. My code like this:

import torch from models import yolo

model = yolo.Model(cfg="models/yolov3.yaml",ch=3, nc=1).to("cpu") weights_path = r"F:\DeepLearning\yolov3-master\yolov3-master\runs\train\exp60\weights\best.pt" checkpcoint = torch.load(weights_path) model.load_state_dict(checkpoint["model"], strict=False) model.eval()

The error informations:

Traceback (most recent call last): File "f:/DeepLearning/yolov3-master/yolov3-master/peppertest.py", line 22, in model.load_state_dict(checkpoint["model"], strict=False) File "D:\anaconda3\lib\site-packages\torch\nn\modules\module.py", line 1379, in load_state_dict state_dict = state_dict.copy() File "D:\anaconda3\lib\site-packages\torch\nn\modules\module.py", line 1130, in getattr raise AttributeError("'{}' object has no attribute '{}'".format( AttributeError: 'Model' object has no attribute 'copy'

Additional

No response

Search before asking

• [X] I have searched the YOLOv3 issues and discussions and found no similar questions.

Question

How can i train yolov4 in this repo sir @glenn-jocher ? I'm new to this field and i need some steps to start transfer learning my data using yolov4 under this repo since i've found some issues that this repo can train yolov4. Can I still do that? How?

Additional

No response

Search before asking

• [X] I have searched the YOLOv3 issues and discussions and found no similar questions.

Question

新的 YOLO4 权重 转换不了 项目提供的CFG 文件 yolo4 darkne 下也无法训练

Additional

No response

Search before asking

• [X] I have searched the YOLOv3 issues and found no similar bug report.

YOLOv3 Component

Detection

Bug

1. Hello,I would like to ask you some questions.Today ,when i used the detect.py of archive branch,i met some errors(the same way used in branch of master is well ):

jetson@jetson-desktop:~/yolov3$ python3 detect.py Namespace(agnostic_nms=False, augment=False, cfg='cfg/yolov3.cfg', classes=None, conf_thres=0.3, device='', fourcc='mp4v', half=False, img_size=512, iou_thres=0.6, names='data/coco.names', output='output', save_txt=False, source='data/samples', view_img=False, weights='weights/yolov3.pt') Using CUDA device0 _CudaDeviceProperties(name='NVIDIA Tegra X2', total_memory=3832MB)

/home/jetson/.local/lib/python3.6/site-packages/torch/functional.py:445: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at /media/nvidia/NVME/pytorch/pytorch-v1.10.0/aten/src/ATen/native/TensorShape.cpp:2157.) return _VF.meshgrid(tensors, **kwargs) # type: ignore[attr-defined] Model Summary: 222 layers, 6.19491e+07 parameters, 6.19491e+07 gradients, 117.5 GFLOPS Traceback (most recent call last): File "detect.py", line 191, in detect() File "detect.py", line 25, in detect model.load_state_dict(torch.load(weights, map_location=device)['model']) File "/home/jetson/.local/lib/python3.6/site-packages/torch/serialization.py", line 607, in load return _load(opened_zipfile, map_location, pickle_module, **pickle_load_args) File "/home/jetson/.local/lib/python3.6/site-packages/torch/serialization.py", line 882, in _load result = unpickler.load() File "/home/jetson/.local/lib/python3.6/site-packages/torch/serialization.py", line 875, in find_class return super().find_class(mod_name, name) ModuleNotFoundError: No module named 'models.yolo'; 'models' is not a package

Environment

No response

Minimal Reproducible Example

No response

Additional

I want to ask some suggestions about how to deploy V3 algorithm on my jetson development board. looking forward to your reply,With best wishes.

Are you willing to submit a PR?

• [X] Yes I'd like to help by submitting a PR!