Ultralytics YOLO Hyperparameter Tuning Guide
Hyperparameter tuning in Ultralytics YOLO is an automated, iterative search that optimizes settings — such as learning rate, loss weights, and augmentation strength — to maximize a machine learning model's performance metrics like accuracy, precision, and recall. Rather than testing these values by hand, Ultralytics YOLO explores the hyperparameter space with a genetic algorithm that mutates and evaluates candidate configurations across many short training runs.
Watch: How to Tune Hyperparameters for Better Model Performance 🚀
What are Hyperparameters?#
Hyperparameters are high-level, structural settings for the algorithm. They are set prior to the training phase and remain constant during it. Here are some commonly tuned hyperparameters in Ultralytics YOLO:
- Learning Rate
lr0: Determines the step size at each iteration while moving towards a minimum in the loss function. - Batch Size
batch: Number of images processed simultaneously in a forward pass. - Number of Epochs
epochs: An epoch is one complete forward and backward pass of all the training examples. - Architecture Specifics: Such as channel counts, number of layers, types of activation functions, etc. These live in the model YAML configuration rather than in training arguments.
Of these four, only the learning rate is in the default search space: batch, epochs, and architecture remain fixed. The tuner also mutates the schedule, loss gains, and augmentations listed below.
Genetic Evolution and Mutation#
Ultralytics YOLO uses genetic algorithms, a family of evolutionary algorithms, to optimize hyperparameters. Genetic algorithms are inspired by the mechanism of natural selection and genetics.
- Selection: Each iteration selects one of up to nine highest-fitness configurations, weighted by fitness. After 30 results, the best 20% of all results (up to 30) also estimate parameter covariance, with influence that grows as elite evidence accumulates.
- Mutation: Each iteration mutates roughly half of the parameters using Gaussian steps normalized to their search ranges. Once covariance is active, a confidence-weighted share of at most 40% uses correlated steps, with reflection returning those steps within bounds without accumulating clipped values. As trials accumulate without improvement, step sizes narrow linearly by up to 20% over 25 trials and reset when a new best result is found. Iteration 1 has no parent and uses the default training hyperparameters as a baseline.
Preparing for Hyperparameter Tuning#
Before you begin the tuning process, it's important to:
- Identify the Metrics: The built-in tuner (
use_ray=False) ranks trials by the task'sfitnessscore, whileuse_ray=Trueuses the task metric. Decide which secondary metrics you will inspect, such as AP50 or F1-score. - Set the Tuning Budget: Define how much compute you're willing to allocate. Hyperparameter tuning can be computationally intensive.
How the Tuning Loop Works#
For each iteration, the built-in tuner repeats the following loop:
- Initialize hyperparameters — start from a reasonable baseline, either the default hyperparameters set by Ultralytics YOLO or values based on your domain knowledge or previous experiments.
- Mutate hyperparameters — the
Tunerclass produces a new set of hyperparameters from the existing set with its_mutatemethod, automatically. - Train the model — train using the mutated hyperparameters, then assess training performance with your chosen metrics.
- Evaluate the model — the evaluation process produces the iteration's fitness score, which is compared against previous iterations to determine whether the current hyperparameters are an improvement.
- Log results — record both the performance metrics and the corresponding hyperparameters for future reference. Ultralytics YOLO automatically saves these results in NDJSON format.
- Repeat — continue until the set number of iterations is reached, with each iteration building on knowledge gained from previous runs. The built-in tuner runs all
iterations; Ray Tune reports metrics after each fit epoch, allowing ASHA to stop weak trials after its grace period.
Iterations and Population Size#
With the built-in tuner (use_ray=False), iterations controls the total number of sequential trials and defaults to 300. Each trial trains one model with one hyperparameter configuration — for example, iterations=40 with epochs=50 schedules 40 independent 50-epoch training runs, not one 50-epoch run with a separate population of 40 candidates.
The built-in genetic algorithm has no explicit population size parameter. Once prior trials exist, it selects from up to nine of the highest-fitness configurations and mutates one candidate per iteration.
For parallel trials or more advanced search strategies, set use_ray=True to use Ray Tune, which uses the same 300-trial default and receives iterations as num_samples. See the Ray Tune integration guide for details.
Default Search Space#
The following table lists the default search space parameters for hyperparameter tuning in YOLO26. Each parameter has a specific value range defined by a tuple (min, max).
| Parameter | Type | Value Range | Description |
|---|---|---|---|
lr0 | float | (1e-5, 1e-2) | Initial learning rate at the start of training. Lower values provide more stable training but slower convergence |
lrf | float | (0.01, 1.0) | Final learning rate factor as a fraction of lr0. Controls how much the learning rate decreases during training |
momentum | float | (0.7, 0.98) | SGD momentum factor. Higher values help maintain consistent gradient direction and can speed up convergence |
weight_decay | float | (0.0, 0.001) | L2 regularization factor to prevent overfitting. Larger values enforce stronger regularization |
warmup_epochs | float | (0.0, 5.0) | Number of epochs for linear learning rate warmup. Helps prevent early training instability |
warmup_momentum | float | (0.0, 0.95) | Initial momentum during warmup phase. Gradually increases to the final momentum value |
box | float | (1.0, 20.0) | Bounding box loss weight in the total loss function. Balances box regression vs classification |
cls | float | (0.1, 4.0) | Classification loss weight in the total loss function. Higher values emphasize correct class prediction |
cls_pw | float | (0.0, 1.0) | Class weighting power for handling class imbalance. Higher values increase weight on rare classes |
dfl | float | (0.4, 12.0) | Weight of the box-distance regression term. DFL-free YOLO26 applies this gain to an L1 loss on normalized box distances; higher values emphasize precise localization |
hsv_h | float | (0.0, 0.1) | Random hue augmentation range in HSV color space. Helps model generalize across color variations |
hsv_s | float | (0.0, 0.9) | Random saturation augmentation range in HSV space. Simulates different lighting conditions |
hsv_v | float | (0.0, 0.9) | Random value (brightness) augmentation range. Helps model handle different exposure levels |
degrees | float | (0.0, 45.0) | Maximum rotation augmentation in degrees. Helps model become invariant to object orientation |
translate | float | (0.0, 0.9) | Maximum translation augmentation as fraction of image size. Improves robustness to object position |
scale | float | (0.0, 0.95) | Random scaling augmentation range. Helps model detect objects at different sizes |
shear | float | (0.0, 10.0) | Maximum shear augmentation in degrees. Adds perspective-like distortions to training images |
perspective | float | (0.0, 0.001) | Random perspective augmentation range. Simulates different viewing angles |
flipud | float | (0.0, 1.0) | Probability of vertical image flip during training. Useful for overhead/aerial imagery |
fliplr | float | (0.0, 1.0) | Probability of horizontal image flip. Helps model become invariant to object direction |
bgr | float | (0.0, 1.0) | Probability of using BGR augmentation, which swaps color channels. Can help with color invariance |
mosaic | float | (0.0, 1.0) | Probability of using mosaic augmentation, which combines 4 images. Especially useful for small object detection |
mixup | float | (0.0, 1.0) | Probability of using mixup augmentation, which blends two images. Can improve model robustness |
cutmix | float | (0.0, 1.0) | Probability of using cutmix augmentation. Combines image regions while maintaining local features |
copy_paste | float | (0.0, 1.0) | In flip mode the fraction of eligible objects copied, in mixup mode also the probability of applying it |
close_mosaic | int | (0.0, 10.0) | Disables mosaic in the last N epochs to stabilize training before completion. Mutated as a float, then rounded to a whole number of epochs |
Custom Search Space Example#
The example below defines a search space and calls model.tune(), which runs the Tuner class, to tune YOLO26n on COCO8 for 30 epochs with the AdamW optimizer. Plotting, checkpointing, and validation other than on the final epoch are skipped for faster tuning.
This example is for demonstration only. Hyperparameters derived from short or small-scale tuning runs are rarely optimal for real-world training. In practice, tuning should be performed under settings similar to full training — including comparable datasets, epochs, and augmentations — to ensure reliable and transferable results. Quick tuning may bias parameters toward faster convergence or short-term validation gains that do not generalize.
from ultralytics import YOLO
# Initialize the YOLO model
model = YOLO("yolo26n.pt")
# Define search space
search_space = {
"lr0": (1e-5, 1e-2),
"degrees": (0.0, 45.0),
}
# Tune hyperparameters on COCO8 for 30 epochs
model.tune(
data="coco8.yaml",
epochs=30,
iterations=300,
optimizer="AdamW",
space=search_space,
plots=False,
save=False,
val=False,
)Resuming an Interrupted Hyperparameter Tuning Session#
You can resume an interrupted hyperparameter tuning session by passing resume=True. You can optionally pass the directory name used under runs/{task} to resume. Without a name, it resumes the directory literally named tune, which is the default name rather than whichever session ran most recently, so pass name for any run you started with a custom one. You also need to provide all the previous training arguments including data, epochs, iterations and space.
from ultralytics import YOLO
# Define a YOLO model
model = YOLO("yolo26n.pt")
# Define search space
search_space = {
"lr0": (1e-5, 1e-2),
"degrees": (0.0, 45.0),
}
# Resume previous run
results = model.tune(data="coco8.yaml", epochs=50, iterations=300, space=search_space, resume=True)
# Resume tuning run with name 'tune_exp'
results = model.tune(data="coco8.yaml", epochs=50, iterations=300, space=search_space, name="tune_exp", resume=True)Results#
After you've successfully completed the hyperparameter tuning process, you will obtain several files and directories that encapsulate the results of the tuning. The following describes each:
File Structure#
Here's what the directory structure of the results will look like. Sweep directories like multitrain/ and multitrain-2/ hold individual tuning iterations, i.e., one model trained per dataset with one set of hyperparameters, in a subdirectory named after each dataset. To save storage, the tuner deletes the per-dataset run directories of iterations that are not the current best. The tune/ directory contains tuning results from all the individual model trainings:
runs/
└── detect/
├── multitrain/
│ └── coco8/
├── multitrain-2/
├── ...
└── tune/
├── best_hyperparameters.yaml
├── tune_fitness.png
├── tune_results.ndjson
├── tune_scatter_plots.png
└── weights/
├── last.pt
└── best.ptFile Descriptions#
best_hyperparameters.yaml#
This YAML file contains the best-performing hyperparameters found during the tuning process. You can use this file to initialize future trainings with these optimized settings.
-
Format: YAML
-
Usage: Hyperparameter results
-
Example:
# 558/900 iterations complete ✅ (45536.81s) # Results saved to /usr/src/ultralytics/runs/detect/tune # Best fitness=0.64297 observed at iteration 498 # Best fitness metrics are {'metrics/precision(B)': 0.87247, 'metrics/recall(B)': 0.71387, 'metrics/mAP50(B)': 0.79106, 'metrics/mAP50-95(B)': 0.62651, 'val/box_loss': 2.79884, 'val/cls_loss': 2.72386, 'val/l1_loss': 0.68503, 'fitness': 0.64297} # Best fitness model is /usr/src/ultralytics/runs/detect/tune/weights # Best fitness hyperparameters are printed below. lr0: 0.00269 lrf: 0.00288 momentum: 0.73375 weight_decay: 0.00015 warmup_epochs: 1.22935 warmup_momentum: 0.1525 box: 18.27875 cls: 1.32899 dfl: 0.56016 hsv_h: 0.01148 hsv_s: 0.53554 hsv_v: 0.13636 degrees: 0.0 translate: 0.12431 scale: 0.07643 shear: 0.0 perspective: 0.0 flipud: 0.0 fliplr: 0.08631 mosaic: 0.42551 mixup: 0.0 copy_paste: 0.0
tune_fitness.png#
This is a plot displaying fitness against the number of iterations. It helps you visualize how the genetic algorithm performed over time.
- Format: PNG
- Usage: Performance visualization
The plot contains:
- One marker per iteration per dataset, so a single-dataset run shows one point per iteration, and a multi-dataset run shows one point per dataset per iteration.
- A dotted "smoothed mean" line computed as a Gaussian smoothing (
sigma=3) over the per-iteration top-level fitness values.
tune_results.ndjson#
An NDJSON file containing detailed results of each tuning iteration. Each line is one JSON object with the aggregate fitness, tuned hyperparameters, and per-dataset metrics. Single-dataset and multi-dataset tuning use the same file format.
- Format: NDJSON
- Usage: Per-iteration results tracking.
- Example:
A pretty-printed example follows for readability; in the actual .ndjson file, each object is stored on a single line.
{
"iteration": 1,
"fitness": 0.48628,
"hyperparameters": {
"lr0": 0.01,
"lrf": 0.01,
"momentum": 0.937,
"weight_decay": 0.0005
},
"datasets": {
"coco8": {
"metrics/precision(B)": 0.65666,
"metrics/recall(B)": 0.85,
"metrics/mAP50(B)": 0.85086,
"metrics/mAP50-95(B)": 0.64104,
"val/box_loss": 1.57958,
"val/cls_loss": 1.04986,
"val/l1_loss": 1.32641,
"fitness": 0.64104
},
"coco8-grayscale": {
"metrics/precision(B)": 0.6582,
"metrics/recall(B)": 0.51667,
"metrics/mAP50(B)": 0.59106,
"metrics/mAP50-95(B)": 0.33152,
"val/box_loss": 1.95424,
"val/cls_loss": 1.64059,
"val/l1_loss": 1.70226,
"fitness": 0.33152
}
},
"save_dirs": {
"coco8": "runs/detect/multitrain/coco8",
"coco8-grayscale": "runs/detect/multitrain/coco8-grayscale"
}
}The top-level fitness is the arithmetic mean of the per-dataset fitness values. For single-dataset tuning the datasets dict has one entry whose fitness equals the top-level fitness. One JSON object is recorded per completed iteration. The actual save_dirs paths are absolute; they are abbreviated above for readability.
tune_scatter_plots.png#
This file contains scatter plots generated from tune_results.ndjson, helping you visualize relationships between different hyperparameters and performance metrics. Mutation uses each hyperparameter's configured range, so values can evolve from zero-valued defaults such as degrees and shear.
- Format: PNG
- Usage: Exploratory data analysis
weights/#
This directory contains the saved PyTorch models from the best-fitness iteration of the tuning run. Both files come from that same iteration:
last.pt: Weights from the last epoch of that iteration's training run.best.pt: Weights from the best-fitness epoch of that iteration's training run.
Using these results, you can make more informed decisions for future model trainings and analyses.
Conclusion#
Hyperparameter tuning in Ultralytics YOLO is both simple to launch and powerful under the hood, using fitness-weighted selection and range-normalized mutation in a genetic algorithm. Following the loop outlined in this guide lets you systematically tune your model for better performance, then reuse the resulting best_hyperparameters.yaml to initialize future training runs. To scale tuning across parallel trials and more advanced search algorithms, continue with the Ray Tune integration guide, or run managed jobs with configurable hyperparameters and real-time metrics tracking on Ultralytics Platform via cloud training.
For settings that no argument exposes, such as per-layer learning rates or a custom fitness metric, subclass the trainer instead. For deeper insights, explore the Tuner class source code. If you have questions or feature requests, reach out on GitHub or Discord.
FAQ#
Set an initial value with the
lr0parameter — common values range from0.001to0.01— and let tuning mutate it from there to find the optimum. You can automate this with themodel.tune()method. For example:Examplefrom ultralytics import YOLO # Initialize the YOLO model model = YOLO("yolo26n.pt") # Tune hyperparameters on COCO8 for 30 epochs model.tune(data="coco8.yaml", epochs=30, iterations=300, optimizer="AdamW", plots=False, save=False, val=False)For more details, check the Ultralytics YOLO configuration page.
Genetic algorithms give Ultralytics YOLO26 a robust way to explore the hyperparameter space without manual trial and error. Key benefits include:
- Efficient Search: Fitness-weighted selection preserves strong configurations while range-normalized mutation explores new candidates.
- Avoiding Local Minima: Random mutations help the search escape local minima that a purely greedy search could get stuck in.
- Performance Metrics: They adapt based on a task-specific fitness score (mAP50-95 for detection).
To see how genetic algorithms can optimize hyperparameters, check out the hyperparameter evolution guide.
The time required for hyperparameter tuning with Ultralytics YOLO largely depends on several factors such as the size of the dataset, the complexity of the model architecture, the number of iterations, and the computational resources available. For instance, tuning YOLO26n on a dataset like COCO8 for 30 epochs might take several hours to days, depending on the hardware.
To effectively manage tuning time, define a clear tuning budget beforehand, as covered in Preparing for Hyperparameter Tuning. This helps balance resource allocation and optimization goals.
When evaluating model performance during hyperparameter tuning in YOLO, you can use several key metrics:
- AP50: The average precision at IoU threshold of 0.50.
- F1-Score: The harmonic mean of precision and recall.
- Precision and Recall: Individual metrics indicating the model's accuracy in identifying true positives versus false positives and false negatives.
These metrics help you understand different aspects of your model's performance.
Yes, Ultralytics YOLO26 integrates with Ray Tune for advanced hyperparameter optimization. Ray Tune offers search algorithms like Bayesian optimization and Hyperband, along with parallel execution capabilities to speed up the tuning process.
To use Ray Tune with YOLO26, set the
use_ray=Trueparameter in yourmodel.tune()method call.
Comments152
Ask a question, share your results, or help someone else.
GengBin414(edited)
I tuned my defect detection dataset ,which are composed of 8000+ images, this is my tune code: model.tune( data="../data/data20260420_line_scan_merged_6ch.yaml", epochs=300, iterations=300, #space=search_space, #optimizer="AdamW", optimizer="MuSGD", lr0=0.01, momentum=0.9, shear=1, degrees=1, perspective=0.0001, mixup=0.1, cutmix=0.1, batch=20 )
I got the best mAP 0.586 at iteration 5 after tuned 30+ iterations, then I retrain the dataset with the following code: model.train( data="../data/data20260420_line_scan_merged_6ch.yaml", epochs=300, batch=20, imgsz=640, device=0, workers=16, optimizer="MuSGD", lr0=0.00721, lrf=0.01267, momentum=0.81816, weight_decay=0.00077, warmup_epochs=3.31429, warmup_momentum=0.95, box=7.11682, cls=0.48318, dfl=0.9397, hsv_h=0.00364, hsv_s=0.9, hsv_v=0.29751, degrees=0.99922, translate=0.07662, scale=0.51303, shear=0.76723, perspective=0.0, flipud=0.00019, fliplr=0.57639, bgr=0.00634, mosaic=0.62424, mixup=0.11307, cutmix=0.13007, copy_paste=0.00583, close_mosaic=10 #cfg="best_hyperparameters_map0.586.yaml" )
I copied the hyperparameters from best_hyperparameters_map0.586.yaml, but I got the best mAP 0.568, what cause the discrepancy?Can sb help me out? Thanks.
Glenn JocherUltralyticsA small gap like
0.586 -> 0.568is usually expected becausetune()saves the best single stochastic run, while your latertrain()is a fresh run with different randomness and augmentation order; please first validate the tuning checkpointruns/detect/tune/weights/best.pt, then retrain with the exact same args plusseed=0anddeterministic=True, and make sure you are comparing the same metric (fitness/mAP50-95) described in the Hyperparameter Tuning guide. If the gap still looks too large, please share the text output from validating that checkpoint and the matching row fromtune_results.ndjsonhere in Discussions so the team can compare them directly.GengBin414(edited)
Thank you for your timely reply, I validated the tuning checkpoint, here is the text output from validating that checkpoint: (yolo26) ps@ps:~/Desktop/GB/ultralytics$ python val.py Ultralytics 8.4.26 🚀 Python-3.11.15 torch-2.5.1 CUDA:0 (NVIDIA GeForce RTX 4090 D, 24208MiB) YOLO26x summary (fused): 190 layers, 55,645,380 parameters, 0 gradients, 193.9 GFLOPs val: Fast image access ✅ (ping: 0.0±0.0 ms, read: 10744.4±1424.6 MB/s, size: 734.1 KB) val: Scanning /home/ps/Desktop/GB/ultralytics/datasets/Merged_6Chanel_20260420/imagesH/val.cache... 786 images, 47 backgrounds, 0 corrupt: 100% ━━━━━━━━━━━━ 786/786 412.1Mit/s 0.0s Class Images Instances Box(P R mAP50 mAP50-95): 100% ━━━━━━━━━━━━ 50/50 6.9it/s 7.2s all 786 1363 0.843 0.69 0.765 0.585 black 45 47 0.914 0.677 0.828 0.648 crack 120 145 0.681 0.517 0.55 0.334 edge 184 195 0.916 0.899 0.94 0.833 line 116 147 0.809 0.646 0.736 0.544 pit 73 89 0.775 0.543 0.642 0.447 raw 279 318 0.861 0.745 0.81 0.644 white 95 106 0.9 0.623 0.712 0.46 yellow 299 316 0.884 0.873 0.9 0.773 Speed: 0.9ms preprocess, 7.3ms inference, 0.0ms loss, 0.0ms postprocess per image Results saved to /home/ps/Desktop/GB/ultralytics/runs/detect/val81
I checked my retrain args, it show that seed=0 and deterministic=True already set: python train.py New https://pypi.org/project/ultralytics/8.4.53 available 😃 Update with 'pip install -U ultralytics' Ultralytics 8.4.26 🚀 Python-3.11.15 torch-2.5.1 CUDA:0 (NVIDIA GeForce RTX 4090 D, 24208MiB) engine/trainer: agnostic_nms=False, amp=True, angle=1.0, augment=False, auto_augment=randaugment, batch=20, bgr=0.00634, box=7.11682, cache=False, cfg=None, classes=None, close_mosaic=10, cls=0.48318, compile=False, conf=None, copy_paste=0.00583, copy_paste_mode=flip, cos_lr=False, cutmix=0.13007, data=../data/data20260420_line_scan_merged_6ch.yaml, degrees=0.99922, deterministic=True, device=0, dfl=0.9397, dnn=False, dropout=0.0, dynamic=False, embed=None, end2end=None, epochs=500, erasing=0.4, exist_ok=False, fliplr=0.57639, flipud=0.00019, format=torchscript, fraction=1.0, freeze=None, half=False, hsv_h=0.00364, hsv_s=0.9, hsv_v=0.29751, imgsz=640, int8=False, iou=0.7, keras=False, kobj=1.0, line_width=None, lr0=0.00721, lrf=0.01267, mask_ratio=4, max_det=300, mixup=0.11307, mode=train, model=yolo26x.pt, momentum=0.81816, mosaic=0.62424, multi_scale=0.0, name=train137, nbs=64, nms=False, opset=None, optimize=False, optimizer=MuSGD, overlap_mask=True, patience=100, perspective=0.0, plots=True, pose=12.0, pretrained=True, profile=False, project=None, rect=False, resume=False, retina_masks=False, rle=1.0, save=True, save_conf=False, save_crop=False, save_dir=/home/ps/Desktop/GB/ultralytics/runs/detect/train137, save_frames=False, save_json=False, save_period=-1, save_txt=False, scale=0.51303, seed=0, shear=0.76723, show=False, show_boxes=True, show_conf=True, show_labels=True, simplify=True, single_cls=False, source=None, split=val, stream_buffer=False, task=detect, time=None, tracker=botsort.yaml, translate=0.07662, val=True, verbose=True, vid_stride=1, visualize=False, warmup_bias_lr=0.1, warmup_epochs=3.31429, warmup_momentum=0.95, weight_decay=0.00077, workers=16, workspace=None
And here is the train output: Epoch GPU_mem box_loss cls_loss dfl_loss Instances Size 260/300 20.8G 0.4503 0.4947 0.004246 50 640: 100% ━━━━━━━━━━━━ 354/354 2.7it/s 2:10 Class Images Instances Box(P R mAP50 mAP50-95): 100% ━━━━━━━━━━━━ 20/20 4.5it/s 4.4s all 786 1363 0.841 0.682 0.751 0.564 EarlyStopping: Training stopped early as no improvement observed in last 100 epochs. Best results observed at epoch 160, best model saved as best.pt. To update EarlyStopping(patience=100) pass a new patience value, i.e.
patience=300or usepatience=0to disable EarlyStopping.260 epochs completed in 9.792 hours. Optimizer stripped from /home/ps/Desktop/GB/ultralytics/runs/detect/train138/weights/last.pt, 118.4MB Optimizer stripped from /home/ps/Desktop/GB/ultralytics/runs/detect/train138/weights/best.pt, 118.4MB
Validating /home/ps/Desktop/GB/ultralytics/runs/detect/train138/weights/best.pt... Ultralytics 8.4.26 🚀 Python-3.11.15 torch-2.5.1 CUDA:0 (NVIDIA GeForce RTX 4090 D, 24208MiB) YOLO26x summary (fused): 190 layers, 55,645,380 parameters, 0 gradients, 193.9 GFLOPs Class Images Instances Box(P R mAP50 mAP50-95): 100% ━━━━━━━━━━━━ 20/20 4.9it/s 4.1s all 786 1363 0.826 0.685 0.758 0.568 black 45 47 0.882 0.634 0.773 0.579 crack 120 145 0.675 0.559 0.577 0.314 edge 184 195 0.925 0.885 0.949 0.824 line 116 147 0.785 0.667 0.714 0.52 pit 73 89 0.84 0.528 0.662 0.447 raw 279 318 0.812 0.72 0.804 0.638 white 95 106 0.802 0.611 0.674 0.434 yellow 299 316 0.888 0.877 0.911 0.787 Speed: 0.1ms preprocess, 4.4ms inference, 0.0ms loss, 0.0ms postprocess per image Results saved to /home/ps/Desktop/GB/ultralytics/runs/detect/train138
And I don't know where to find tune_results.ndjson mentioned in your reply. I guess you mean the matching row from tune_results.csv then here is: fitness lr0 lrf momentum weight_decay warmup_epochs warmup_momentum box cls dfl hsv_h hsv_s hsv_v degrees translate scale shear perspective flipud fliplr bgr mosaic mixup cutmix copy_paste close_mosaic 0.58582 0.00721 0.01267 0.81816 0.00077 3.31429 0.95 7.11682 0.48318 0.9397 0.00364 0.9 0.29751 0.99922 0.07662 0.51303 0.76723 0 0.00019 0.57639 0.00634 0.62424 0.11307 0.13007 0.00583 10
Glenn JocherUltralyticsThe main issue I see is that your retrain is not matching the tuned trial: the runtime log shows
epochs=500rather than300, so the schedule driven bylr0andlrfis different and that alone can explain0.585 -> 0.568; the current Hyperparameter Tuning guide writestune_results.ndjsonunderruns/detect/tune/, but in your older8.4.26install the equivalent may betune_results.csv, so please rerun with the exact runtime args from the best trial, especiallyepochs=300and the samepatience, and if it still printsepochs=500then another override is being applied.GengBin414(edited)
Sorry, after hyperparameter tuning, I retrained first with epochs=500, observed mAP=0.57, then I changed epochs=300, observed mAP=0.568. In the previous reply, I mistakenly copied the retrain args from the epochs=500 run. But the train output is correctly copied from the epochs=300 run, you can tell this out from the EarlyStopping output: Epoch GPU_mem box_loss cls_loss dfl_loss Instances Size 260/300 20.8G 0.4503 0.4947 0.004246 50 640: 100% ━━━━━━━━━━━━ 354/354 2.7it/s 2:10 Class Images Instances Box(P R mAP50 mAP50-95): 100% ━━━━━━━━━━━━ 20/20 4.5it/s 4.4s all 786 1363 0.841 0.682 0.751 0.564 EarlyStopping: Training stopped early as no improvement observed in last 100 epochs. Best results observed at epoch 160, best model saved as best.pt. To update EarlyStopping(patience=100) pass a new patience value, i.e. patience=300 or use patience=0 to disable EarlyStopping.
And the retrain args of the epochs=300 run, it shows that seed=0 and deterministic=True already set: python train.py New https://pypi.org/project/ultralytics/8.4.53 available 😃 Update with 'pip install -U ultralytics' Ultralytics 8.4.26 🚀 Python-3.11.15 torch-2.5.1 CUDA:0 (NVIDIA GeForce RTX 4090 D, 24208MiB) engine/trainer: agnostic_nms=False, amp=True, angle=1.0, augment=False, auto_augment=randaugment, batch=20, bgr=0.00634, box=7.11682, cache=False, cfg=None, classes=None, close_mosaic=10, cls=0.48318, compile=False, conf=None, copy_paste=0.00583, copy_paste_mode=flip, cos_lr=False, cutmix=0.13007, data=../data/data20260420_line_scan_merged_6ch.yaml, degrees=0.99922, deterministic=True, device=0, dfl=0.9397, dnn=False, dropout=0.0, dynamic=False, embed=None, end2end=None, epochs=300, erasing=0.4, exist_ok=False, fliplr=0.57639, flipud=0.00019, format=torchscript, fraction=1.0, freeze=None, half=False, hsv_h=0.00364, hsv_s=0.9, hsv_v=0.29751, imgsz=640, int8=False, iou=0.7, keras=False, kobj=1.0, line_width=None, lr0=0.00721, lrf=0.01267, mask_ratio=4, max_det=300, mixup=0.11307, mode=train, model=yolo26x.pt, momentum=0.81816, mosaic=0.62424, multi_scale=0.0, name=train138, nbs=64, nms=False, opset=None, optimize=False, optimizer=MuSGD, overlap_mask=True, patience=100, perspective=0.0, plots=True, pose=12.0, pretrained=True, profile=False, project=None, rect=False, resume=False, retina_masks=False, rle=1.0, save=True, save_conf=False, save_crop=False, save_dir=/home/ps/Desktop/GB/ultralytics/runs/detect/train138, save_frames=False, save_json=False, save_period=-1, save_txt=False, scale=0.51303, seed=0, shear=0.76723, show=False, show_boxes=True, show_conf=True, show_labels=True, simplify=True, single_cls=False, source=None, split=val, stream_buffer=False, task=detect, time=None, tracker=botsort.yaml, translate=0.07662, val=True, verbose=True, vid_stride=1, visualize=False, warmup_bias_lr=0.1, warmup_epochs=3.31429, warmup_momentum=0.95, weight_decay=0.00077, workers=16, workspace=None
Murat Raimbekovquick check on what is likely going on, framed generally: tune saves the best of N stochastic trials, and a single retrain is one new stochastic draw with the same hyperparameters. bit-exact reproduction between the two is not guaranteed even with
seed=0, deterministic=True. ourinit_seeds(deterministic=True)setstorch.use_deterministic_algorithms(True, warn_only=True)inultralytics/utils/torch_utils.py, matching the pytorch reproducibility notes - cuda ops without a deterministic alternative still run non-deterministically with just a warning. over many epochs on a real dataset that variance can compound.a few practical takeaways before assuming the gap is a tuning bug:
- your trial weights at
runs/detect/tune/weights/best.ptalready validate to 0.585, so the trial result is real and reusable - if you need exactly that performance for deployment, use that checkpoint directly. - the recommended way to reuse tuned hyperparameters in a fresh train run is
cfg="runs/detect/tune/best_hyperparameters.yaml"(the commented-out line in your script). that file contains the exact values the trial used and matches the workflow described in the hyperparameter tuning guide. - to bound run-to-run variance, retrain a few times with
seed=0,seed=1,seed=2and compare the mean against the tune's reported best, not a single retrain.
unrelated, you are on ultralytics 8.4.26 and the current release is 8.4.53; for new tuning runs it is worth updating.
- your trial weights at
UserHi, I have a question about Hyperparameter Tuning. What value of
epochsanditerationsshould I pass to model.tune(). Default value of epochs is 100, and iterations is 10. Are the default values OK, or should I change them ? My dataset is about 10,000 images.
Glenn JocherUltralyticsFor a 10k-image dataset, the defaults are fine for a quick sanity check, but for useful tuning you should set
epochsclose to your real training length becauseepochsis per trial, whileiterationsis how many different trials are tested; a practical starting point isepochs=50to100withiterations=30to100depending on your compute budget, then retrain once with the best settings frombest_hyperparameters.yamlas shown in the Hyperparameter Tuning guide.1
UserThank you very much ! I will try to search the best hyperparameters !
Glenn JocherUltralyticsYou're welcome — sounds good; if anything looks off during tuning, please share your
model.tune()call plus the resultingbest_hyperparameters.yamlortune_results.ndjsonhere, and we can help compare it against the expected outputs in the Hyperparameter Tuning guide.
harunemrekiran61
You’ve shown the box range as 0.02 to 0.2, but the default value appears to be 7.5. Is there a mistake here? Could you please check?
Glenn JocherUltralyticsGood catch —
boxis a loss gain (default7.5during training), so the0.02–0.2range shown in the docs table is outdated/incorrect; the current built-in tuner boundsboxto1.0–20.0(see the Tuner class reference), and you can always override it viaspace={...}as shown in the Hyperparameter Tuning guide.
Larry-zhai
How can I directly apply the optimized best_hyperparameters.yaml to my training? In fact, when I use code similar to the following: from ultralytics import YOLO model = YOLO("yolov8n.pt") model.train(data="data.yaml", hyp="hyp.yaml") an error occurs: SyntaxError: 'hyp' is not a valid YOLO argument.
J.Q. PanHi I have used the above code to apply the tuning but neither the diagram nor the models will be saved. I only get a single tune folder with all tuning details. I may ask if someone else has also the same problem.
BrianNieuwenhuis
When running the code below the model.tune() call does not recognize my model specification. Could someone explain to me what I am doing wrong. I am using model.overrides so that I can also tune the shear and perspective parameters that are otherwise initialized at 0.
from ultralytics import YOLO import wandb from ultralytics import settings settings.update({"wandb": True})
import argparse
def main(): # parse command-line args if needed parser = argparse.ArgumentParser() parser.add_argument("--gpus", type=int, default=1) parser.add_argument("--workers", type=int, default=4) args = parser.parse_args()
# Define a YOLO model model = YOLO("yolo11x.pt") # Dynamically assign GPUs based on --gpus argument device = list(range(args.gpus)) # e.g., [0, 1, 2, 3] if args.gpus = 4 # Define search space search_space = { "lr0": (1e-5, 1e-1), "lrf": (0.01, 1.0), "momentum": (0.6, 0.98), "weight_decay": (0.0, 0.001), "warmup_epochs": (0, 5), "warmup_momentum": (0.0,0.95), "box": (0.02, 0.2), "cls": (0.2, 4.0), "hsv_h": (0.0, 0.1), "hsv_s": (0.0, 0.9), "hsv_v": (0.0, 0.9), "translate":(0.0, 0.9), "scale": (0.0, 0.9), "shear": (0.0, 10.0), "perspective": (0.0, 0.001), "mosaic": (0.0,1.0), "mixup": (0.0, 1.0), "copy_paste": (0.0, 1.0) } # Manually initialize certain hperparameters that would not be tuned otherwise model.overrides = { 'shear': 1, 'perspective': 0.001, 'mixup': 0.01, 'copy_paste' : 0.01, } # tune the model results = model.tune(project='./tune/crop1344',name="tune_default_batch16_v100", use_ray = False, data="data.yaml", #hyp="custom_hyp.yaml", # this does not seem to work work for tune scripts!! batch = 16, imgsz = 1344, device = device, workers = args.workers, degrees = 180, #these parameters are fixed in the model tune flipud = 0.5, fliplr = 0.5, bgr = 0, auto_augment = None, #erasing = 0, epochs = 100, iterations=10, space=search_space, optimizer="AdamW", resume=False, #start over or resume previous checkpoint plots=False, # don't make plots etc to save time until last timepoint save=False, val=False, verbose=False)if name == "main": main()
yyagmurkyaa
Hello again, I have trained new models using the parameters you provided and by organizing my datasets. However, a common issue across all of them is that the models mistakenly classify objects like myself or nearby computer screens as the traffic sign "turn right ahead" (which is one of the labeled classes in my dataset). I have checked my dataset and confirmed that there are no such unrelated objects in the background.
Additionally, the models sometimes confuse left and right directional signs.
Why might these errors be occurring, and how can I fix them? Thank you in advance for your help.
Glenn JocherUltralytics(edited)@Fatima-khanjar to balance precision and recall in your over-predicting model, try these adjustments:
- Increase confidence threshold during inference (not training) to filter out low-confidence predictions
- Adjust IoU threshold in NMS to reduce duplicate detections
- Increase classification loss weight with
cls=1.0or higher to improve class discrimination - Check annotation quality in your dataset, especially for small or ambiguous objects
- Add hard negative examples to your dataset - images with objects similar to your classes but not actually part of them
These changes should help improve precision without significantly sacrificing your excellent recall. For implementation details, see our NMS documentation for confidence thresholding examples.
@yyagmurkyaa for your traffic sign confusion issues:
-
For the model detecting unrelated objects as "turn right ahead" signs:
- Add "hard negative" examples to your training set (images with computers, people, but no traffic signs)
- Increase
cls=1.0or higher to improve classification accuracy - Check if the "turn right ahead" sign has visual similarities to these false positives
-
For left/right confusion:
- Create specific augmentations for directional signs without flipping (as you noted)
- Ensure balanced examples of both directional types
- Consider using a slightly larger model like YOLO26m for better feature discrimination
These targeted adjustments should help address the specific confusion patterns you're seeing.
Fatima-khanjar
Hello again!!
First of all, thank you so much for your help @glenn-jocher __ you've truly helped me a lot, and I really appreciate it!
I have a question regarding fine-tuning a pre-trained YOLO model (e.g., trained on the COCO dataset) for my custom use case. My dataset includes a combination of both new classes (not present in COCO) and some of the existing COCO classes , but with different layouts or contexts.
What I’d like to do is the following:
Add my new classes starting from index 80, while still leveraging the pre-trained weights for the existing COCO classes.
Ideally, I want to preserve the learned knowledge for the COCO classes (especially the ones shared with my dataset), and only train the model further on the new classes.
Is it possible to:
Fine-tune the model in such a way that it uses the pre-trained weights for the common classes, and only learns new classes from scratch starting from a different index?
Retain and benefit from the original COCO knowledge while introducing and training on new class definitions?
Also, I’ve noticed that when I fine-tune a pre-trained model and define my own data.yaml file — completely different from coco.yaml — I redefine the class names and indexes based on my custom dataset, completely removing the COCO classes not present in my data. In this case, what really happens during fine-tuning or transfer learning under the hood? Is the model discarding unused class weights? Are shared class weights still used effectively?
I'd greatly appreciate any clarification or guidance on how to approach this setup properly.
Best regards,
Glenn JocherUltralytics@Fatima-khanjar great question about fine-tuning with mixed COCO and custom classes!
Yes, you can leverage pre-trained COCO weights while adding new classes. Here's how:
-
When fine-tuning a YOLO model, the model's detection head is automatically resized to accommodate your new class count. The weights for existing classes are preserved where possible.
-
For classes that overlap with COCO, the model will leverage transfer learning from those pre-trained weights, while new classes will be initialized with random weights.
-
You don't need to explicitly start from index 80 - just define your classes in your data.yaml file, and the model will handle the mapping automatically.
-
During fine-tuning, the model doesn't discard unused class weights but rather reshapes the classification head to match your class count. Weights for classes that exist in both datasets are transferred and further optimized for your specific data.
For best results with mixed classes, I recommend using a smaller learning rate initially (e.g.,
lr0=0.001) to preserve the knowledge in the pre-trained weights while allowing the model to adapt to your new classes.-
Fatima-khanjar
Hello! @glenn-jocher thank you so much for your explanation!
I have a follow-up question regarding class index matching. Let's say the person class in the COCO dataset is indexed as 0, but in my data.yaml file, I assign person to index 5. In this case, the model will not treat the person at index 5 as the same class as COCO’s index 0, correct?
So, to benefit from the pretrained weights of shared classes, do the class indices in my custom dataset need to match the COCO indices exactly?
Best regards,
Glenn JocherUltralytics(edited)@Fatima-khanjar you're asking a crucial question about class mapping during transfer learning!
You no longer need to keep COCO's class indices. With
cls_remap=True(the default), training copies the pretrained classification-head rows by class name, so a class such aspersonkeeps its pretrained weights at whatever index it has in your dataset, while new classes such astraffic_signstart from fresh weights. Training printsRemapped N/M cls head rows from pretrained weights by class namewhen rows are copied, andcls_remap=Falsedisables it. For example, thisdata.yamlworks:names: 0: person 1: car 2: traffic_signJust make sure the shared classes use the same names as in COCO. See the Fine-Tuning Guide for details.
Fatima-khanjar
@glenn-jocher Hello again!!!
Thank you so much for your clarification! I have a quick follow-up question.
Let’s say the COCO dataset defines the following class indices in its YAML file: 0: person, 2: car, 3: cat.
In my own dataset, I share only the person and car classes with COCO — I don’t want to include cat. If I define my shared classes using the same indices as COCO (i.e., 0: person, 2: car), but then replace the COCO class cat at index 3 with a new class (e.g., 3: animal), will I still benefit from the pretrained weights for the shared classes (person and car)?
Or will assigning a new class at an index previously used by COCO (like 3) cause any issues?
best regards,
Glenn JocherUltralytics(edited)@Fatima-khanjar great question about reusing COCO indices for new classes!
With
cls_remap=True(the default), pretrained classification-head rows are matched by class name, not by index. Shared classes such aspersonandcarkeep their pretrained weights wherever they sit in your dataset. A new class like "animal" at index 3 does not inherit the pretrained "cat" row, because the names do not match, so it starts from fresh weights. Reusing a COCO index therefore neither helps nor hurts the shared classes.If a new class is close to a pretrained one, give it that class's name so its row is copied. Using a lower learning rate initially (e.g.,
lr0=0.001) also helps preserve the pretrained features while the model adapts to your new classes.1
yyagmurkyaa
Thank you for your feedback. I trained my dataset using YOLOv11s. Although I didn’t experience severe overfitting with the current parameters, I couldn’t achieve optimal performance either. Some traffic signs were misclassified or not detected at all.
While searching for the best hyperparameters, I realized that I hadn't applied data augmentation—except for the flip parameter. I have an example parameter file, and I plan to include all relevant augmentation options except for flipping.
I would really appreciate your help in finding the most suitable parameters for my dataset, as I’m planning to participate in a competition soon. I need to enhance my model by adjusting the parameters properly to achieve the best possible performance.
Thank you in advance for your support.
On Sun, Apr 20, 2025 at 1:51 PM phatima khanjar @.***> wrote:
Hello,
I hope this message finds you well.
image.png (view on web) https://github.com/user-attachments/assets/be6db84b-fe0d-4356-ba77-9b4426c253ee
I fine-tuned YOLOv11s on my custom dataset, which is relatively small (1,156 images split into 75% training, 15% validation, and 10% test). The dataset consists of 41 classes, with each class containing between 200 and 300 images. I experimented with three different sets of hyperparameters: the default values, those optimized using Ray Tune, and others obtained through a Genetic Algorithm. In all cases, I observed oscillations in the validation dfl and box losses, which may indicate signs of overfitting. I would appreciate any suggestions or insights on how to address this issue, particularly given the small size of the dataset, and how to improve model generalization.
best regards,
— Reply to this email directly, view it on GitHub https://github.com/orgs/ultralytics/discussions/9536#discussioncomment-12889695, or unsubscribe https://github.com/notifications/unsubscribe-auth/BPQ7ZLAQ2G7ZL434FN3WFYD22N345AVCNFSM6AAAAABFVXPSL2VHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTEOBYHE3DSNI . You are receiving this because you were mentioned.Message ID: @.*** com>
Glenn JocherUltralytics(edited)@yyagmurkyaa for traffic sign detection and classification with YOLO26s, I recommend focusing on these specific augmentation and training parameters:
from ultralytics import YOLO model = YOLO("yolo26s.pt") model.train( data="your_traffic_signs.yaml", epochs=100, patience=25, batch=16, imgsz=640, # Consider 800+ for small traffic signs hsv_h=0.015, # Color augmentation helps with varying lighting conditions hsv_s=0.7, hsv_v=0.4, scale=0.5, # Scale augmentation critical for varying distances translate=0.1, # Helps with signs at different positions mosaic=1.0, # Excellent for object detection tasks mixup=0.1, # Light mixup can help with generalization perspective=0.0005, # Slight perspective change for different viewing angles weight_decay=0.0005, # Regularization to reduce overfitting optimizer="AdamW", lr0=0.001, lrf=0.01, cos_lr=True # Cosine learning rate scheduler )For traffic sign detection specifically:
- Increase image size to better detect small signs
- Focus on color-preserving augmentations as colors are critical for traffic signs
- Use perspective transformations to simulate different viewing angles
- Enable strong scale/translate augmentations to handle varying distances
Consider running
model.tune()with these parameters as starting points to find the optimal configuration for your competition. You might also need to address class imbalance if some signs appear less frequently in your dataset.yyagmurkyaa
You're very welcome! I wasn't expecting such a quick response either. I'm excited to hear that you'll be training with these parameters and look forward to seeing the results as soon as possible. Thank you again for your support!
Glenn JocherUltralytics(edited)@Fatima-khanjar for a small dataset (1,156 images) with 41 classes, oscillations in validation losses often indicate overfitting. Here are specific recommendations to improve generalization:
-
Strengthen regularization:
- Increase
weight_decayto 0.001-0.003 - Add dropout with
dropout=0.1to reduce co-adaptation
- Increase
-
Enhance data augmentation:
- Maximize augmentations:
mosaic=1.0,mixup=0.2,copy_paste=0.1 - Use strong geometric transforms:
translate=0.2,scale=0.5
- Maximize augmentations:
-
Training modifications:
- Reduce model complexity by using YOLO26n instead of YOLO26s
- Implement gradual unfreezing (
freezelayers at first) - Enable
cos_lr=Truefor smoother learning rate decay
-
Consider transfer learning modifications:
- Start with a model pre-trained on a domain similar to yours
- Use fewer epochs with higher
patiencefor early stopping
With 41 classes and limited samples per class, these techniques should help stabilize training and improve generalization. You might also benefit from training ensembles of multiple models if computational resources allow. For more details, refer to our Model Training Tips.
-
Fatima-khanjar(edited)
Hello,
Thank you for your advice! Your instructions have improved my model’s performance compared to before. I have now achieved the following metrics:
Mean Precision: 0.794 Mean Recall: 0.99 Mean F1 Score: 0.881 mAP@0.5: 0.938 mAP@0.5-0.95: 0.827
After analyzing these results, it seems that the model is over-predicting. Do you have any suggestions on how I can better balance precision and recall?
best regards,
Fatima-khanjar
Hello,
I hope this message finds you well.
I fine-tuned YOLOv11s on my custom dataset, which is relatively small (1,156 images split into 75% training, 15% validation, and 10% test). The dataset consists of 41 classes, with each class containing between 200 and 300 images. I experimented with three different sets of hyperparameters: the default values, those optimized using Ray Tune, and others obtained through a Genetic Algorithm. In all cases, I observed oscillations in the validation dfl and box losses, which may indicate signs of overfitting. I would appreciate any suggestions or insights on how to address this issue, particularly given the small size of the dataset, and how to improve model generalization.
best regards,
yyagmurkyaa
Hello, I am training a model using YOLOv11, and the dataset consists of traffic signs. There are 22 classes, and each class has 400 labeled images. I’m not sure what parameters I should use for training. What parameters should I set to prevent overfitting, and what are the correct values? Could you help me with this?
Glenn JocherUltralytics(edited)@yyagmurkyaa your dataset of 22 classes with 400 images per class should provide a good foundation for traffic sign detection. To prevent overfitting with YOLO26, I'd recommend:
- Start with hyperparameter tuning using the
model.tune()method to find optimal values:
from ultralytics import YOLO model = YOLO("yolo26n.pt") # Start with a smaller model for faster tuning model.tune( data="your_traffic_signs.yaml", epochs=30, iterations=100, weight_decay=0.0005, # Try higher values like 0.001-0.005 to increase regularization patience=15, # Early stopping val=True )-
Key hyperparameters to prevent overfitting:
- Increase
weight_decay(0.001-0.005) - Enable strong data augmentation (
hsv_h,hsv_s,translate,scale,fliplr,mosaic) - Use early stopping with reasonable
patience - Consider reducing model capacity (use n or s model variants)
- Increase
-
Monitor validation metrics to detect overfitting (when training accuracy continues improving but validation metrics plateau or decline)
The automated tuning process should help identify the optimal parameter values for your specific dataset. For more details, check out our Hyperparameter Tuning Guide.
- Start with hyperparameter tuning using the
yyagmurkyaa
Thank you for your feedback. I trained my dataset using YOLOv11s. Although I didn’t experience severe overfitting with the current parameters, I couldn’t achieve optimal performance either. Some traffic signs were misclassified or not detected at all.
While searching for the best hyperparameters, I realized that I hadn't applied data augmentation—except for the flip parameter. I have an example parameter file, and I plan to include all relevant augmentation options except for flipping.
I would really appreciate your help in finding the most suitable parameters for my dataset, as I’m planning to participate in a competition soon. I need to enhance my model by adjusting the parameters properly to achieve the best possible performance.
Thank you in advance for your support.