Double Labeling Problem

sezer-muhammed commented 3 years ago

Hi, As can you see below, some objects labeled twice by model. NMS prevents double label if the labels are same, but for different labels it does not prevent.

I looked the code and change merge_nms vallue to True, but it does not help.

How can I prevent that? I need to apply nms to my class 0, 1, 2, 3, 4, 5, 6 together, and class 7 is different than them and does not need to nms with them. But it does not matter, how can ı apply nms to all of them together?

glenn-jocher commented 3 years ago

@sezer-muhammed you can turn agnostic NMS on, or you can turn multi-label NMS off, both of which will remove multiple classes labelled on the same instance. A bit more information is below:

👋 Hello, thank you for asking about the differences between train.py, detect.py and test.py in YOLOv5.

These 3 files are designed for different purposes and utilize different dataloaders with different settings. train.py dataloaders are designed for a speed-accuracy compromise, test.py is designed to obtain the best mAP on a validation dataset, and detect.py is designed for best real-world inference results. A few important aspects of each:

train.py

trainloader LoadImagesAndLabels(): designed to load train dataset images and labels. Augmentation capable and enabled. https://github.com/ultralytics/yolov5/blob/fca5e2a48fb526b57bda0c66be6b7ac1aaa8d83d/train.py#L188-L192
testloader LoadImagesAndLabels(): designed to load val dataset images and labels. Augmentation capable but disabled. https://github.com/ultralytics/yolov5/blob/fca5e2a48fb526b57bda0c66be6b7ac1aaa8d83d/train.py#L199-L202
image size: 640
confidence threshold: 0.001
iou threshold: 0.6
multi-label: True
padding: None

test.py

dataloader LoadImagesAndLabels(): designed to load train, val, test dataset images and labels. Augmentation capable but disabled. https://github.com/ultralytics/yolov5/blob/fca5e2a48fb526b57bda0c66be6b7ac1aaa8d83d/test.py#L89-L90
image size: 640
confidence threshold: 0.001
iou threshold: 0.6
multi-label: True
padding: 0.5 * maximum stride

detect.py

dataloaders (multiple): designed for loading multiple types of media (images, videos, globs, directories, streams). https://github.com/ultralytics/yolov5/blob/fca5e2a48fb526b57bda0c66be6b7ac1aaa8d83d/detect.py#L46-L53
image size: 640
confidence threshold: 0.25
iou threshold: 0.45
multi-label: False
padding: None

YOLOv5 PyTorch Hub

autoShape() class used for image loading, preprocessing, inference and NMS. For more info see YOLOv5 PyTorch Hub Tutorial https://github.com/ultralytics/yolov5/blob/fca5e2a48fb526b57bda0c66be6b7ac1aaa8d83d/models/common.py#L225-L250

sezer-muhammed commented 3 years ago

@sezer-muhammed you can turn agnostic NMS on, or you can turn multi-label NMS off, both of which will remove multiple classes labelled on the same instance. A bit more information is below:

👋 Hello, thank you for asking about the differences between train.py, detect.py and test.py in YOLOv5.

These 3 files are designed for different purposes and utilize different dataloaders with different settings. train.py dataloaders are designed for a speed-accuracy compromise, test.py is designed to obtain the best mAP on a validation dataset, and detect.py is designed for best real-world inference results. A few important aspects of each:

train.py

trainloader LoadImagesAndLabels(): designed to load train dataset images and labels. Augmentation capable and enabled. https://github.com/ultralytics/yolov5/blob/fca5e2a48fb526b57bda0c66be6b7ac1aaa8d83d/train.py#L188-L192

testloader LoadImagesAndLabels(): designed to load val dataset images and labels. Augmentation capable but disabled. https://github.com/ultralytics/yolov5/blob/fca5e2a48fb526b57bda0c66be6b7ac1aaa8d83d/train.py#L199-L202

image size: 640

confidence threshold: 0.001

iou threshold: 0.6

multi-label: True

padding: None

test.py

dataloader LoadImagesAndLabels(): designed to load train, val, test dataset images and labels. Augmentation capable but disabled. https://github.com/ultralytics/yolov5/blob/fca5e2a48fb526b57bda0c66be6b7ac1aaa8d83d/test.py#L89-L90

image size: 640

confidence threshold: 0.001

iou threshold: 0.6

multi-label: True

padding: 0.5 * maximum stride

detect.py

dataloaders (multiple): designed for loading multiple types of media (images, videos, globs, directories, streams). https://github.com/ultralytics/yolov5/blob/fca5e2a48fb526b57bda0c66be6b7ac1aaa8d83d/detect.py#L46-L53

image size: 640

confidence threshold: 0.25

iou threshold: 0.45

multi-label: False

padding: None

YOLOv5 PyTorch Hub

autoShape() class used for image loading, preprocessing, inference and NMS. For more info see YOLOv5 PyTorch Hub Tutorial https://github.com/ultralytics/yolov5/blob/fca5e2a48fb526b57bda0c66be6b7ac1aaa8d83d/models/common.py#L225-L250

Hi again,

I tried turning multiple label off it did not work. now i turned on agnostic variable, did not work

and i turned on agnostic and tund off multiple label and here is the result :(

glenn-jocher commented 3 years ago

@sezer-muhammed you've incorrectly applied the above advice. With agnostic NMS in place the above image you are showing is not produceable.

sezer-muhammed commented 3 years ago

@sezer-muhammed you've incorrectly applied the above advice. With agnostic NMS in place the above image you are showing is not producable.

I will look again, I changed codes a lot maybe i have broken something. so, I wont close this issue for now.

Thanks!

glenn-jocher commented 3 years ago

@sezer-muhammed yes try with a clean git clone. Agnostic NMS passes all detections through NMS and treats them as a single class, so there can be no inter-class overlaps like what you show.

sezer-muhammed commented 3 years ago

@sezer-muhammed yes try with a clean git clone. Agnostic NMS passes all detections through NMS and treats them as a single class, so there can be no inter-class overlaps like what you show.

DONE! There was two codes that I wrote. I was changing original one but not the one I used...... :(

Ankit-Vohra commented 3 years ago

Can I choose a particular class if there's an inter-class overlapping? Moreover, my model is working a bit weird. I'm training an Object detector for 2 classes (Good and Defective) and it's detecting the object as both the classes with confidence 0.7 and 0.8. Is there any way through which I can improve on the training @glenn-jocher

glenn-jocher commented 3 years ago

@Ankit-Vohra 👋 Hello! Thanks for asking about improving YOLOv5 🚀 training results.

Most of the time good results can be obtained with no changes to the models or training settings, provided your dataset is sufficiently large and well labelled. If at first you don't get good results, there are steps you might be able to take to improve, but we always recommend users first train with all default settings before considering any changes. This helps establish a performance baseline and spot areas for improvement.

If you have questions about your training results we recommend you provide the maximum amount of information possible if you expect a helpful response, including results plots (train losses, val losses, P, R, mAP), PR curve, confusion matrix, training mosaics, test results and dataset statistics images such as labels.png. All of these are located in your project/name directory, typically yolov5/runs/train/exp.

We've put together a full guide for users looking to get the best results on their YOLOv5 trainings below.

Dataset

Images per class. ≥ 1500 images per class recommended
Instances per class. ≥ 10000 instances (labeled objects) per class recommended
Image variety. Must be representative of deployed environment. For real-world use cases we recommend images from different times of day, different seasons, different weather, different lighting, different angles, different sources (scraped online, collected locally, different cameras) etc.
Label consistency. All instances of all classes in all images must be labelled. Partial labelling will not work.
Label accuracy. Labels must closely enclose each object. No space should exist between an object and it's bounding box. No objects should be missing a label.
Background images. Background images are images with no objects that are added to a dataset to reduce False Positives (FP). We recommend about 0-10% background images to help reduce FPs (COCO has 1000 background images for reference, 1% of the total). No labels are required for background images.

Model Selection

Larger models like YOLOv5x and YOLOv5x6 will produce better results in nearly all cases, but have more parameters, require more CUDA memory to train, and are slower to run. For mobile deployments we recommend YOLOv5s/m, for cloud deployments we recommend YOLOv5l/x. See our README table for a full comparison of all models.

YOLOv5 Models

Start from Pretrained weights. Recommended for small to medium sized datasets (i.e. VOC, VisDrone, GlobalWheat). Pass the name of the model to the --weights argument. Models download automatically from the latest YOLOv5 release.

python train.py --data custom.yaml --weights yolov5s.pt
                                         yolov5m.pt
                                         yolov5l.pt
                                         yolov5x.pt
                                         custom_pretrained.pt

Start from Scratch. Recommended for large datasets (i.e. COCO, Objects365, OIv6). Pass the model architecture yaml you are interested in, along with an empty --weights '' argument:

python train.py --data custom.yaml --weights '' --cfg yolov5s.yaml
                                                  yolov5m.yaml
                                                  yolov5l.yaml
                                                  yolov5x.yaml

Training Settings

Before modifying anything, first train with default settings to establish a performance baseline. A full list of train.py settings can be found in the train.py argparser.

Epochs. Start with 300 epochs. If this overfits early then you can reduce epochs. If overfitting does not occur after 300 epochs, train longer, i.e. 600, 1200 etc epochs.
Image size. COCO trains at native resolution of --img 640, though due to the high amount of small objects in the dataset it can benefit from training at higher resolutions such as --img 1280. If there are many small objects then custom datasets will benefit from training at native or higher resolution. Best inference results are obtained at the same --img as the training was run at, i.e. if you train at --img 1280 you should also test and detect at --img 1280.
Batch size. Use the largest --batch-size that your hardware allows for. Small batch sizes produce poor batchnorm statistics and should be avoided.
Hyperparameters. Default hyperparameters are in hyp.scratch.yaml. We recommend you train with default hyperparameters first before thinking of modifying any. In general, increasing augmentation hyperparameters will reduce and delay overfitting, allowing for longer trainings and higher final mAP. Reduction in loss component gain hyperparameters like hyp['obj'] will help reduce overfitting in those specific loss components. For an automated method of optimizing these hyperparameters, see our Hyperparameter Evolution Tutorial.

ultralytics / yolov5