deep_learning/ultralytics
1
0
Fork 0
mirror of https://github.com/ultralytics/ultralytics.git synced 2025-09-15 15:48:41 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

docs: improved documentation for yoloe export (#21728)

Browse Source 
Co-authored-by: UltralyticsAssistant <web@ultralytics.com>
Co-authored-by: Jing Qiu <61612323+Laughing-q@users.noreply.github.com>
This commit is contained in:
Alexis Schutzger 2025-08-19 08:37:41 +02:00 committed by GitHub
parent cc8783e2d0
commit bc85202ca6
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 99 additions and 38 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

135
docs/en/models/yoloe.md
View File

@ -89,23 +89,16 @@ You can fine-tune any [pretrained YOLOE model](#textvisual-prompt-models) on you
```python ```python
from ultralytics import YOLOE from ultralytics import YOLOE
from ultralytics.models.yolo.yoloe import YOLOEPESegTrainer as Trainer from ultralytics.models.yolo.yoloe import YOLOEPESegTrainer
model = YOLOE("yoloe-11s-seg.pt") model = YOLOE("yoloe-11s-seg.pt")
# Fine-tune on your segmentation dataset
results = model.train( results = model.train(
data="coco128-seg.yaml", data="coco128-seg.yaml", # Segmentation dataset
epochs=80, epochs=80,
close_mosaic=10, patience=10,
batch=16, trainer=YOLOEPESegTrainer, # <- Important: use segmentation trainer
optimizer="AdamW",
lr0=1e-3,
warmup_bias_lr=0.0,
weight_decay=0.025,
momentum=0.9,
workers=4,
device="0",
trainer=Trainer,
) )
``` ```
@ -115,11 +108,21 @@ You can fine-tune any [pretrained YOLOE model](#textvisual-prompt-models) on you
```python ```python
from ultralytics import YOLOE from ultralytics import YOLOE
from ultralytics.models.yolo.yoloe import YOLOEPETrainer as Trainer # noqa from ultralytics.models.yolo.yoloe import YOLOEPETrainer
# create detection model from yaml then load segmentation weights # Initialize a detection model from a config
model = YOLOE("yoloe-11s.yaml").load("yoloe-11s-seg.pt") model = YOLOE("yoloe-11s.yaml")
# rest of the code is same as the instance segmentation fine-tuning example above
# Load weights from a pretrained segmentation checkpoint (same scale)
model.load("yoloe-11s-seg.pt")
# Fine-tune on your detection dataset
results = model.train(
data="coco128.yaml", # Detection dataset
epochs=80,
patience=10,
trainer=YOLOEPETrainer, # <- Important: use detection trainer
)
``` ```
=== "Linear Probing" === "Linear Probing"
@ -130,13 +133,18 @@ You can fine-tune any [pretrained YOLOE model](#textvisual-prompt-models) on you
```python ```python
from ultralytics import YOLOE from ultralytics import YOLOE
from ultralytics.models.yolo.yoloe import YOLOEPESegTrainer as Trainer from ultralytics.models.yolo.yoloe import YOLOEPESegTrainer
# Load a pretrained segmentation model
model = YOLOE("yoloe-11s-seg.pt") model = YOLOE("yoloe-11s-seg.pt")
head_index = len(model.model.model) - 1
freeze = [str(f) for f in range(0, head_index)] # freeze all layers except head
# Freeze all head components except classification branch # Identify the head layer index
head_index = len(model.model.model) - 1
# Freeze all backbone and neck layers (i.e. everything before the head)
freeze = [str(i) for i in range(0, head_index)]
# Freeze parts of the segmentation head, keeping only the classification branch trainable
for name, child in model.model.model[-1].named_children(): for name, child in model.model.model[-1].named_children():
if "cv3" not in name: if "cv3" not in name:
freeze.append(f"{head_index}.{name}") freeze.append(f"{head_index}.{name}")
@ -153,19 +161,12 @@ You can fine-tune any [pretrained YOLOE model](#textvisual-prompt-models) on you
] ]
) )
# Train only the classification branch
results = model.train( results = model.train(
data="coco128-seg.yaml", data="coco128-seg.yaml", # Segmentation dataset
epochs=2, epochs=80,
close_mosaic=0, patience=10,
batch=16, trainer=YOLOEPESegTrainer, # <- Important: use segmentation trainer
optimizer="AdamW",
lr0=1e-3,
warmup_bias_lr=0.0,
weight_decay=0.025,
momentum=0.9,
workers=4,
device="0",
trainer=Trainer,
freeze=freeze, freeze=freeze,
) )
``` ```
@ -176,11 +177,45 @@ You can fine-tune any [pretrained YOLOE model](#textvisual-prompt-models) on you
```python ```python
from ultralytics import YOLOE from ultralytics import YOLOE
from ultralytics.models.yolo.yoloe import YOLOEPETrainer as Trainer # noqa from ultralytics.models.yolo.yoloe import YOLOEPETrainer
# create detection model from yaml then load segmentation weights # Initialize a detection model from a config
model = YOLOE("yoloe-11s.yaml").load("yoloe-11s-seg.pt") model = YOLOE("yoloe-11s.yaml")
# rest of the code is same as the instance segmentation fine-tuning example above
# Load weights from a pretrained segmentation checkpoint (same scale)
model.load("yoloe-11s-seg.pt")
# Identify the head layer index
head_index = len(model.model.model) - 1
# Freeze all backbone and neck layers (i.e. everything before the head)
freeze = [str(i) for i in range(0, head_index)]
# Freeze parts of the segmentation head, keeping only the classification branch trainable
for name, child in model.model.model[-1].named_children():
if "cv3" not in name:
freeze.append(f"{head_index}.{name}")
# Freeze detection branch components
freeze.extend(
[
f"{head_index}.cv3.0.0",
f"{head_index}.cv3.0.1",
f"{head_index}.cv3.1.0",
f"{head_index}.cv3.1.1",
f"{head_index}.cv3.2.0",
f"{head_index}.cv3.2.1",
]
)
# Train only the classification branch
results = model.train(
data="coco128.yaml", # Detection dataset
epochs=80,
patience=10,
trainer=YOLOEPETrainer, # <- Important: use detection trainer
freeze=freeze,
)
``` ```
### Predict Usage ### Predict Usage
@ -363,6 +398,8 @@ YOLOE supports both text-based and visual prompting. Using prompts is straightfo
### Val Usage ### Val Usage
Model validation on a dataset is streamlined as follows:
!!! example !!! example
=== "Text Prompt" === "Text Prompt"
@ -417,7 +454,31 @@ YOLOE supports both text-based and visual prompting. Using prompts is straightfo
metrics = model.val(data="coco128-seg.yaml") metrics = model.val(data="coco128-seg.yaml")
``` ```
Model validation on a dataset is streamlined as follows: ### Export Usage
The export process is similar to other YOLO models, with the added flexibility of handling text and visual prompts:
!!! example
```python
from ultralytics import YOLOE
# Select yoloe-11s/m-seg.pt for different sizes
model = YOLOE("yoloe-11l-seg.pt")
# Configure the set_classes() before exporting the model
names = ["person", "bus"]
model.set_classes(names, model.get_text_pe(names))
export_model = model.export(format="onnx")
model = YOLOE(export_model)
# Run detection on the given image
results = model.predict("path/to/image.jpg")
# Show results
results[0].show()
```
### Train Official Models ### Train Official Models

2
docs/en/tasks/classify.md
View File

@ -87,7 +87,7 @@ Train YOLO11n-cls on the MNIST160 dataset for 100 [epochs](https://www.ultralyti
from ultralytics import YOLO from ultralytics import YOLO
from ultralytics.data.dataset import ClassificationDataset from ultralytics.data.dataset import ClassificationDataset
from ultralytics.models.yolo.classify import ClassificationTrainer from ultralytics.models.yolo.classify import ClassificationTrainer, ClassificationValidator
class CustomizedDataset(ClassificationDataset): class CustomizedDataset(ClassificationDataset):