are the layers count valid for YOLOv11-seg models?

[pranta-barua007]

Variant	Backbone Modules	Backbone Layers	Neck Modules	Neck Layers	Head Modules	Head Layers	Total Layers
n (nano)	11	36	12	28	1	139	203
s (small)	11	36	12	28	1	139	203
m (medium)	11	36	12	28	1	189	253
l (large)	11	41	12	30	1	308	379
x (x-large)	11	41	12	30	1	308	379

[UltralyticsAssistant]

👋 Hello @pranta-barua007, thank you for your interest in Ultralytics 🚀! We recommend a visit to the Docs for new users where you can find many Python and CLI usage examples and where many of the most common questions may already be answered.

If this is a 🐛 Bug Report, please provide a minimum reproducible example to help us debug it.

If you have a custom training ❓ Question, please provide as much detail as possible, such as dataset examples, training logs, and confirm you are following our Tips for Best Training Results.

Join the Ultralytics community where it suits you best! For real-time chat, head to Discord 🎧. Prefer in-depth discussions? Check out Discourse. Or dive into threads on our Subreddit to share and learn from others.

Upgrade

Upgrade to the latest ultralytics package including all requirements in a Python>=3.8 environment with PyTorch>=1.8 to verify your issue is not already resolved in the latest version:

pip install -U ultralytics

Environments

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

• 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 Ultralytics CI tests are currently passing. CI tests verify correct operation of all YOLO Modes and Tasks on macOS, Windows, and Ubuntu every 24 hours and on every commit.

---

This is an automated response to help streamline support 😊. An Ultralytics engineer will review and assist you here soon!

[pranta-barua007]

here is information of models

INFO:
YOLO11n-seg summary: 203 layers, 2,876,848 parameters, 0 gradients, 10.5 GFLOPs
(203, 2876848, 0, 10.5294976)
YOLO11s-seg summary: 203 layers, 10,113,248 parameters, 0 gradients, 35.8 GFLOPs
(203, 10113248, 0, 35.7621248)
YOLO11m-seg summary: 253 layers, 22,420,896 parameters, 0 gradients, 123.9 GFLOPs
(253, 22420896, 0, 123.91982079999998)
YOLO11l-seg summary: 379 layers, 27,678,368 parameters, 0 gradients, 143.0 GFLOPs
(379, 27678368, 0, 143.004928)
YOLO11x-seg summary: 379 layers, 62,142,656 parameters, 0 gradients, 320.2 GFLOPs
(379, 62142656, 0, 320.1852928)

YAML file for all variants:
{'nc': 80,
'scales': {'n': [0.5, 0.25, 1024],
's': [0.5, 0.5, 1024],
'm': [0.5, 1.0, 512],
'l': [1.0, 1.0, 512],
'x': [1.0, 1.5, 512]},
'backbone': [[-1, 1, 'Conv', [64, 3, 2]],
[-1, 1, 'Conv', [128, 3, 2]],
[-1, 2, 'C3k2', [256, False, 0.25]],
[-1, 1, 'Conv', [256, 3, 2]],
[-1, 2, 'C3k2', [512, False, 0.25]],
[-1, 1, 'Conv', [512, 3, 2]],
[-1, 2, 'C3k2', [512, True]],
[-1, 1, 'Conv', [1024, 3, 2]],
[-1, 2, 'C3k2', [1024, True]],
[-1, 1, 'SPPF', [1024, 5]],
[-1, 2, 'C2PSA', [1024]]],
'head': [[-1, 1, 'nn.Upsample', ['None', 2, 'nearest']],
[[-1, 6], 1, 'Concat', [1]],
[-1, 2, 'C3k2', [512, False]],
[-1, 1, 'nn.Upsample', ['None', 2, 'nearest']],
[[-1, 4], 1, 'Concat', [1]],
[-1, 2, 'C3k2', [256, False]],
[-1, 1, 'Conv', [256, 3, 2]],
[[-1, 13], 1, 'Concat', [1]],
[-1, 2, 'C3k2', [512, False]],
[-1, 1, 'Conv', [512, 3, 2]],
[[-1, 10], 1, 'Concat', [1]],
[-1, 2, 'C3k2', [1024, True]],
[[16, 19, 22], 1, 'Segment', ['nc', 32, 256]]],
'scale': 'n',
'yaml_file': 'yolo11n-seg.yaml',
'ch': 3}
{'nc': 80,
'scales': {'n': [0.5, 0.25, 1024],
's': [0.5, 0.5, 1024],
'm': [0.5, 1.0, 512],
'l': [1.0, 1.0, 512],
'x': [1.0, 1.5, 512]},
'backbone': [[-1, 1, 'Conv', [64, 3, 2]],
[-1, 1, 'Conv', [128, 3, 2]],
[-1, 2, 'C3k2', [256, False, 0.25]],
[-1, 1, 'Conv', [256, 3, 2]],
[-1, 2, 'C3k2', [512, False, 0.25]],
[-1, 1, 'Conv', [512, 3, 2]],
[-1, 2, 'C3k2', [512, True]],
[-1, 1, 'Conv', [1024, 3, 2]],
[-1, 2, 'C3k2', [1024, True]],
[-1, 1, 'SPPF', [1024, 5]],
[-1, 2, 'C2PSA', [1024]]],
'head': [[-1, 1, 'nn.Upsample', ['None', 2, 'nearest']],
[[-1, 6], 1, 'Concat', [1]],
[-1, 2, 'C3k2', [512, False]],
[-1, 1, 'nn.Upsample', ['None', 2, 'nearest']],
[[-1, 4], 1, 'Concat', [1]],
[-1, 2, 'C3k2', [256, False]],
[-1, 1, 'Conv', [256, 3, 2]],
[[-1, 13], 1, 'Concat', [1]],
[-1, 2, 'C3k2', [512, False]],
[-1, 1, 'Conv', [512, 3, 2]],
[[-1, 10], 1, 'Concat', [1]],
[-1, 2, 'C3k2', [1024, True]],
[[16, 19, 22], 1, 'Segment', ['nc', 32, 256]]],
'scale': 's',
'yaml_file': 'yolo11s-seg.yaml',
'ch': 3}
{'nc': 80,
'scales': {'n': [0.5, 0.25, 1024],
's': [0.5, 0.5, 1024],
'm': [0.5, 1.0, 512],
'l': [1.0, 1.0, 512],
'x': [1.0, 1.5, 512]},
'backbone': [[-1, 1, 'Conv', [64, 3, 2]],
[-1, 1, 'Conv', [128, 3, 2]],
[-1, 2, 'C3k2', [256, False, 0.25]],
[-1, 1, 'Conv', [256, 3, 2]],
[-1, 2, 'C3k2', [512, False, 0.25]],
[-1, 1, 'Conv', [512, 3, 2]],
[-1, 2, 'C3k2', [512, True]],
[-1, 1, 'Conv', [1024, 3, 2]],
[-1, 2, 'C3k2', [1024, True]],
[-1, 1, 'SPPF', [1024, 5]],
[-1, 2, 'C2PSA', [1024]]],
'head': [[-1, 1, 'nn.Upsample', ['None', 2, 'nearest']],
[[-1, 6], 1, 'Concat', [1]],
[-1, 2, 'C3k2', [512, False]],
[-1, 1, 'nn.Upsample', ['None', 2, 'nearest']],
[[-1, 4], 1, 'Concat', [1]],
[-1, 2, 'C3k2', [256, False]],
[-1, 1, 'Conv', [256, 3, 2]],
[[-1, 13], 1, 'Concat', [1]],
[-1, 2, 'C3k2', [512, False]],
[-1, 1, 'Conv', [512, 3, 2]],
[[-1, 10], 1, 'Concat', [1]],
[-1, 2, 'C3k2', [1024, True]],
[[16, 19, 22], 1, 'Segment', ['nc', 32, 256]]],
'scale': 'm',
'yaml_file': 'yolo11m-seg.yaml',
'ch': 3}
{'nc': 80,
'scales': {'n': [0.5, 0.25, 1024],
's': [0.5, 0.5, 1024],
'm': [0.5, 1.0, 512],
'l': [1.0, 1.0, 512],
'x': [1.0, 1.5, 512]},
'backbone': [[-1, 1, 'Conv', [64, 3, 2]],
[-1, 1, 'Conv', [128, 3, 2]],
[-1, 2, 'C3k2', [256, False, 0.25]],
[-1, 1, 'Conv', [256, 3, 2]],
[-1, 2, 'C3k2', [512, False, 0.25]],
[-1, 1, 'Conv', [512, 3, 2]],
[-1, 2, 'C3k2', [512, True]],
[-1, 1, 'Conv', [1024, 3, 2]],
[-1, 2, 'C3k2', [1024, True]],
[-1, 1, 'SPPF', [1024, 5]],
[-1, 2, 'C2PSA', [1024]]],
'head': [[-1, 1, 'nn.Upsample', ['None', 2, 'nearest']],
[[-1, 6], 1, 'Concat', [1]],
[-1, 2, 'C3k2', [512, False]],
[-1, 1, 'nn.Upsample', ['None', 2, 'nearest']],
[[-1, 4], 1, 'Concat', [1]],
[-1, 2, 'C3k2', [256, False]],
[-1, 1, 'Conv', [256, 3, 2]],
[[-1, 13], 1, 'Concat', [1]],
[-1, 2, 'C3k2', [512, False]],
[-1, 1, 'Conv', [512, 3, 2]],
[[-1, 10], 1, 'Concat', [1]],
[-1, 2, 'C3k2', [1024, True]],
[[16, 19, 22], 1, 'Segment', ['nc', 32, 256]]],
'scale': 'l',
'yaml_file': 'yolo11l-seg.yaml',
'ch': 3}
{'nc': 80,
'scales': {'n': [0.5, 0.25, 1024],
's': [0.5, 0.5, 1024],
'm': [0.5, 1.0, 512],
'l': [1.0, 1.0, 512],
'x': [1.0, 1.5, 512]},
'backbone': [[-1, 1, 'Conv', [64, 3, 2]],
[-1, 1, 'Conv', [128, 3, 2]],
[-1, 2, 'C3k2', [256, False, 0.25]],
[-1, 1, 'Conv', [256, 3, 2]],
[-1, 2, 'C3k2', [512, False, 0.25]],
[-1, 1, 'Conv', [512, 3, 2]],
[-1, 2, 'C3k2', [512, True]],
[-1, 1, 'Conv', [1024, 3, 2]],
[-1, 2, 'C3k2', [1024, True]],
[-1, 1, 'SPPF', [1024, 5]],
[-1, 2, 'C2PSA', [1024]]],
'head': [[-1, 1, 'nn.Upsample', ['None', 2, 'nearest']],
[[-1, 6], 1, 'Concat', [1]],
[-1, 2, 'C3k2', [512, False]],
[-1, 1, 'nn.Upsample', ['None', 2, 'nearest']],
[[-1, 4], 1, 'Concat', [1]],
[-1, 2, 'C3k2', [256, False]],
[-1, 1, 'Conv', [256, 3, 2]],
[[-1, 13], 1, 'Concat', [1]],
[-1, 2, 'C3k2', [512, False]],
[-1, 1, 'Conv', [512, 3, 2]],
[[-1, 10], 1, 'Concat', [1]],
[-1, 2, 'C3k2', [1024, True]],
[[16, 19, 22], 1, 'Segment', ['nc', 32, 256]]],
'scale': 'x',
'yaml_file': 'yolo11x-seg.yaml',
'ch': 3}

[glenn-jocher]

Perfect! Your model.info() output confirms that your original layer count table was indeed accurate. The layer counts match exactly:

• n/s: 203 layers
• m: 253 layers
• l/x: 379 layers

The differences occur because the 'scales' parameter controls both depth_multiple and width_multiple scaling. Notice how n/s/m use depth_multiple=0.5 while l/x use depth_multiple=1.0, which increases the number of repeated modules in blocks like C3k2 and C2PSA, resulting in more layers for the larger variants.

Thanks for sharing the detailed breakdown - it's a great reference for understanding YOLO11-seg architecture scaling!

[pranta-barua007]

@glenn-jocher now i have actually make this table to validate two things

1. while doing feature extraction if i want to freeze all backbone layers so then if i set freeze=11 (https://docs.ultralytics.com/modes/train/#train-settings ) while training will it freeze the whole backbone of the model keeping only neck and head ?

2. and i also want to validate my intuation on we actully pass module count in freeze right not the layers count?

[glenn-jocher]

Exactly right on both points!

1. Yes, setting freeze=11 will freeze the entire backbone since your YAML shows exactly 11 backbone modules across all YOLO11-seg variants, leaving only the neck and head trainable.

2. Correct - the freeze parameter uses module count, not layer count. This is why you can use the same freeze=11 value across all model variants (n/s/m/l/x) despite them having different total layer counts (203/253/379). The module structure remains consistent, just scaled differently in width/depth.

Your table is a great reference for understanding exactly which modules get frozen at different freeze values!

[pranta-barua007]

@glenn-jocher thanks for clarification but the docs are kinda misleading right? it should be denoted that 'freeze' params accept module count as input not the layer count?