Chiebot-Mirror
/
ultralytics
mirror of https://github.com/ultralytics/ultralytics.git
9
0
Fork 1
Code Issues Packages Projects Releases Wiki Activity

Update OpenVINO INT8 export (#7515)

Browse Source 
Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>
Co-authored-by: UltralyticsAssistant <web@ultralytics.com>
pull/5280/head^2
Glenn Jocher 11 months ago committed by GitHub
parent 4dc8c406f9
commit 30185e0d4c
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
16 changed files with 42 additions and 30 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 2
      docs/en/datasets/explorer/index.md
  2. 1
      docs/en/datasets/index.md
  3. 2
      docs/en/modes/benchmark.md
  4. 2
      docs/en/modes/export.md
  5. 2
      docs/en/modes/predict.md
  6. 2
      docs/en/modes/track.md
  7. 1
      docs/en/reference/data/explorer/gui/dash.md
  8. 2
      docs/en/tasks/classify.md
  9. 2
      docs/en/tasks/detect.md
  10. 16
      docs/en/tasks/obb.md
  11. 2
      docs/en/tasks/pose.md
  12. 2
      docs/en/tasks/segment.md
  13. 2
      docs/en/usage/cli.md
  14. 1
      docs/en/usage/python.md
  15. 20
      docs/en/yolov5/tutorials/model_export.md
  16. 13
      ultralytics/engine/exporter.py

2
docs/en/datasets/explorer/index.md
Unescape View File

@ -25,7 +25,6 @@ pip install ultralytics[explorer]
Explorer works on embedding/semantic search & SQL querying and is powered by [LanceDB](https://lancedb.com/) serverless vector database. Unlike traditional in-memory DBs, it is persisted on disk without sacrificing performance, so you can scale locally to large datasets like COCO without running out of memory.
### Explorer API
This is a Python API for Exploring your datasets. It also powers the GUI Explorer. You can use this to create your own exploratory notebooks or scripts to get insights into your datasets.
@ -41,6 +40,7 @@ yolo explorer
```
!!! note "Note"
Ask AI feature works using OpenAI, so you'll be prompted to set the api key for OpenAI when you first run the GUI.
You can set it like this - `yolo settings openai_api_key="..."`

1
docs/en/datasets/index.md
Unescape View File

@ -16,7 +16,6 @@ Create embeddings for your dataset, search for similar images, run SQL queries,
<img width="1709" alt="Screenshot 2024-01-08 at 7 19 48 PM (1)" src="https://github.com/AyushExel/assets/assets/15766192/e536b0eb-6bce-43fe-b800-3e79510d2e5b">
</p>
- Try the [GUI Demo](explorer/index.md)
- Learn more about the [Explorer API](explorer/index.md)

2
docs/en/modes/benchmark.md
Unescape View File

@ -91,7 +91,7 @@ Benchmarks will attempt to run automatically on all possible export formats belo
| [PyTorch](https://pytorch.org/) | - | `yolov8n.pt` | ✅ | - |
| [TorchScript](https://pytorch.org/docs/stable/jit.html) | `torchscript` | `yolov8n.torchscript` | ✅ | `imgsz`, `optimize` |
| [ONNX](https://onnx.ai/) | `onnx` | `yolov8n.onnx` | ✅ | `imgsz`, `half`, `dynamic`, `simplify`, `opset` |
| [OpenVINO](https://docs.openvino.ai/latest/index.html) | `openvino` | `yolov8n_openvino_model/` | ✅ | `imgsz`, `half`, `int8` |
| [OpenVINO](../integrations/openvino.md) | `openvino` | `yolov8n_openvino_model/` | ✅ | `imgsz`, `half`, `int8` |
| [TensorRT](https://developer.nvidia.com/tensorrt) | `engine` | `yolov8n.engine` | ✅ | `imgsz`, `half`, `dynamic`, `simplify`, `workspace` |
| [CoreML](https://github.com/apple/coremltools) | `coreml` | `yolov8n.mlpackage` | ✅ | `imgsz`, `half`, `int8`, `nms` |
| [TF SavedModel](https://www.tensorflow.org/guide/saved_model) | `saved_model` | `yolov8n_saved_model/` | ✅ | `imgsz`, `keras`, `int8` |

2
docs/en/modes/export.md
Unescape View File

@ -96,7 +96,7 @@ Available YOLOv8 export formats are in the table below. You can export to any fo
| [PyTorch](https://pytorch.org/) | - | `yolov8n.pt` | ✅ | - |
| [TorchScript](https://pytorch.org/docs/stable/jit.html) | `torchscript` | `yolov8n.torchscript` | ✅ | `imgsz`, `optimize` |
| [ONNX](https://onnx.ai/) | `onnx` | `yolov8n.onnx` | ✅ | `imgsz`, `half`, `dynamic`, `simplify`, `opset` |
| [OpenVINO](https://docs.openvino.ai/latest/index.html) | `openvino` | `yolov8n_openvino_model/` | ✅ | `imgsz`, `half`, `int8` |
| [OpenVINO](../integrations/openvino.md) | `openvino` | `yolov8n_openvino_model/` | ✅ | `imgsz`, `half`, `int8` |
| [TensorRT](https://developer.nvidia.com/tensorrt) | `engine` | `yolov8n.engine` | ✅ | `imgsz`, `half`, `dynamic`, `simplify`, `workspace` |
| [CoreML](https://github.com/apple/coremltools) | `coreml` | `yolov8n.mlpackage` | ✅ | `imgsz`, `half`, `int8`, `nms` |
| [TF SavedModel](https://www.tensorflow.org/guide/saved_model) | `saved_model` | `yolov8n_saved_model/` | ✅ | `imgsz`, `keras`, `int8` |

2
docs/en/modes/predict.md
Unescape View File

@ -761,5 +761,7 @@ Here's a Python script using OpenCV (`cv2`) and YOLOv8 to run inference on video
This script will run predictions on each frame of the video, visualize the results, and display them in a window. The loop can be exited by pressing 'q'.
[car spare parts]: https://github.com/RizwanMunawar/ultralytics/assets/62513924/a0f802a8-0776-44cf-8f17-93974a4a28a1
[football player detect]: https://github.com/RizwanMunawar/ultralytics/assets/62513924/7d320e1f-fc57-4d7f-a691-78ee579c3442
[human fall detect]: https://github.com/RizwanMunawar/ultralytics/assets/62513924/86437c4a-3227-4eee-90ef-9efb697bdb43

2
docs/en/modes/track.md
Unescape View File

@ -354,5 +354,7 @@ To initiate your contribution, please refer to our [Contributing Guide](https://
Together, let's enhance the tracking capabilities of the Ultralytics YOLO ecosystem 🙏!
[fish track]: https://github.com/RizwanMunawar/ultralytics/assets/62513924/a5146d0f-bfa8-4e0a-b7df-3c1446cd8142
[people track]: https://github.com/RizwanMunawar/ultralytics/assets/62513924/93bb4ee2-77a0-4e4e-8eb6-eb8f527f0527
[vehicle track]: https://github.com/RizwanMunawar/ultralytics/assets/62513924/ee6e6038-383b-4f21-ac29-b2a1c7d386ab

1
docs/en/reference/data/explorer/gui/dash.md
Unescape View File

@ -4,7 +4,6 @@ description: Detailed reference for the Explorer GUI. Includes brief description
keywords: Ultralytics, data explorer, gui, function reference, documentation, AI queries, image similarity, SQL queries, streamlit, semantic search
---
# Reference for `ultralytics/data/explorer/gui/dash.py`
!!! Note

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

@ -167,7 +167,7 @@ Available YOLOv8-cls export formats are in the table below. You can predict or v
| [PyTorch](https://pytorch.org/) | - | `yolov8n-cls.pt` | ✅ | - |
| [TorchScript](https://pytorch.org/docs/stable/jit.html) | `torchscript` | `yolov8n-cls.torchscript` | ✅ | `imgsz`, `optimize` |
| [ONNX](https://onnx.ai/) | `onnx` | `yolov8n-cls.onnx` | ✅ | `imgsz`, `half`, `dynamic`, `simplify`, `opset` |
| [OpenVINO](https://docs.openvino.ai/latest/index.html) | `openvino` | `yolov8n-cls_openvino_model/` | ✅ | `imgsz`, `half` |
| [OpenVINO](../integrations/openvino.md) | `openvino` | `yolov8n-cls_openvino_model/` | ✅ | `imgsz`, `half`, `int8` |
| [TensorRT](https://developer.nvidia.com/tensorrt) | `engine` | `yolov8n-cls.engine` | ✅ | `imgsz`, `half`, `dynamic`, `simplify`, `workspace` |
| [CoreML](https://github.com/apple/coremltools) | `coreml` | `yolov8n-cls.mlpackage` | ✅ | `imgsz`, `half`, `int8`, `nms` |
| [TF SavedModel](https://www.tensorflow.org/guide/saved_model) | `saved_model` | `yolov8n-cls_saved_model/` | ✅ | `imgsz`, `keras` |

2
docs/en/tasks/detect.md
Unescape View File

@ -168,7 +168,7 @@ Available YOLOv8 export formats are in the table below. You can predict or valid
| [PyTorch](https://pytorch.org/) | - | `yolov8n.pt` | ✅ | - |
| [TorchScript](https://pytorch.org/docs/stable/jit.html) | `torchscript` | `yolov8n.torchscript` | ✅ | `imgsz`, `optimize` |
| [ONNX](https://onnx.ai/) | `onnx` | `yolov8n.onnx` | ✅ | `imgsz`, `half`, `dynamic`, `simplify`, `opset` |
| [OpenVINO](https://docs.openvino.ai/latest/index.html) | `openvino` | `yolov8n_openvino_model/` | ✅ | `imgsz`, `half`, `int8` |
| [OpenVINO](../integrations/openvino.md) | `openvino` | `yolov8n_openvino_model/` | ✅ | `imgsz`, `half`, `int8` |
| [TensorRT](https://developer.nvidia.com/tensorrt) | `engine` | `yolov8n.engine` | ✅ | `imgsz`, `half`, `dynamic`, `simplify`, `workspace` |
| [CoreML](https://github.com/apple/coremltools) | `coreml` | `yolov8n.mlpackage` | ✅ | `imgsz`, `half`, `int8`, `nms` |
| [TF SavedModel](https://www.tensorflow.org/guide/saved_model) | `saved_model` | `yolov8n_saved_model/` | ✅ | `imgsz`, `keras`, `int8` |

16
docs/en/tasks/obb.md
Unescape View File

@ -29,13 +29,13 @@ YOLOv8 pretrained OBB models are shown here, which are pretrained on the [DOTAv1
[Models](https://github.com/ultralytics/ultralytics/tree/main/ultralytics/cfg/models) download automatically from the latest Ultralytics [release](https://github.com/ultralytics/assets/releases) on first use.
| Model | size<br><sup>(pixels) | mAP<sup>test<br>50 | Speed<br><sup>CPU ONNX<br>(ms) | Speed<br><sup>A100 TensorRT<br>(ms) | params<br><sup>(M) | FLOPs<br><sup>(B) |
|----------------------------------------------------------------------------------------------|-----------------------| -------------------- | -------------------------------- | ------------------------------------- | -------------------- | ----------------- |
| [YOLOv8n-obb](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8n-obb.pt) | 1024 | 76.9 | 204.77 | 3.57 | 3.1 | 23.3 |
| [YOLOv8s-obb](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8s-obb.pt) | 1024 | 78.0 | 424.88 | 4.07 | 11.4 | 76.3 |
| [YOLOv8m-obb](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8m-obb.pt) | 1024 | 80.5 | 763.48 | 7.61 | 26.4 | 208.6 |
| [YOLOv8l-obb](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8l-obb.pt) | 1024 | 80.7 | 1278.42 | 11.83 | 44.5 | 433.8 |
| [YOLOv8x-obb](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8x-obb.pt) | 1024 | 81.36 | 1759.10 | 13.23 | 69.5 | 676.7 |
| Model | size<br><sup>(pixels) | mAP<sup>test<br>50 | Speed<br><sup>CPU ONNX<br>(ms) | Speed<br><sup>A100 TensorRT<br>(ms) | params<br><sup>(M) | FLOPs<br><sup>(B) |
|----------------------------------------------------------------------------------------------|-----------------------|--------------------|--------------------------------|-------------------------------------|--------------------|-------------------|
| [YOLOv8n-obb](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8n-obb.pt) | 1024 | 76.9 | 204.77 | 3.57 | 3.1 | 23.3 |
| [YOLOv8s-obb](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8s-obb.pt) | 1024 | 78.0 | 424.88 | 4.07 | 11.4 | 76.3 |
| [YOLOv8m-obb](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8m-obb.pt) | 1024 | 80.5 | 763.48 | 7.61 | 26.4 | 208.6 |
| [YOLOv8l-obb](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8l-obb.pt) | 1024 | 80.7 | 1278.42 | 11.83 | 44.5 | 433.8 |
| [YOLOv8x-obb](https://github.com/ultralytics/assets/releases/download/v8.1.0/yolov8x-obb.pt) | 1024 | 81.36 | 1759.10 | 13.23 | 69.5 | 676.7 |
- **mAP<sup>test</sup>** values are for single-model multi-scale on [DOTAv1 test](https://captain-whu.github.io/DOTA/index.html) dataset. <br>Reproduce by `yolo val obb data=DOTAv1.yaml device=0 split=test` and submit merged results to [DOTA evaluation](https://captain-whu.github.io/DOTA/evaluation.html).
- **Speed** averaged over DOTAv1 val images using an [Amazon EC2 P4d](https://aws.amazon.com/ec2/instance-types/p4/) instance. <br>Reproduce by `yolo val obb data=DOTAv1.yaml batch=1 device=0|cpu`
@ -165,7 +165,7 @@ Available YOLOv8-obb export formats are in the table below. You can predict or v
| [PyTorch](https://pytorch.org/) | - | `yolov8n-obb.pt` | ✅ | - |
| [TorchScript](https://pytorch.org/docs/stable/jit.html) | `torchscript` | `yolov8n-obb.torchscript` | ✅ | `imgsz`, `optimize` |
| [ONNX](https://onnx.ai/) | `onnx` | `yolov8n-obb.onnx` | ✅ | `imgsz`, `half`, `dynamic`, `simplify`, `opset` |
| [OpenVINO](https://docs.openvino.ai/latest/index.html) | `openvino` | `yolov8n-obb_openvino_model/` | ✅ | `imgsz`, `half` |
| [OpenVINO](../integrations/openvino.md) | `openvino` | `yolov8n-obb_openvino_model/` | ✅ | `imgsz`, `half`, `int8` |
| [TensorRT](https://developer.nvidia.com/tensorrt) | `engine` | `yolov8n-obb.engine` | ✅ | `imgsz`, `half`, `dynamic`, `simplify`, `workspace` |
| [CoreML](https://github.com/apple/coremltools) | `coreml` | `yolov8n-obb.mlpackage` | ✅ | `imgsz`, `half`, `int8`, `nms` |
| [TF SavedModel](https://www.tensorflow.org/guide/saved_model) | `saved_model` | `yolov8n-obb_saved_model/` | ✅ | `imgsz`, `keras` |

2
docs/en/tasks/pose.md
Unescape View File

@ -170,7 +170,7 @@ Available YOLOv8-pose export formats are in the table below. You can predict or
| [PyTorch](https://pytorch.org/) | - | `yolov8n-pose.pt` | ✅ | - |
| [TorchScript](https://pytorch.org/docs/stable/jit.html) | `torchscript` | `yolov8n-pose.torchscript` | ✅ | `imgsz`, `optimize` |
| [ONNX](https://onnx.ai/) | `onnx` | `yolov8n-pose.onnx` | ✅ | `imgsz`, `half`, `dynamic`, `simplify`, `opset` |
| [OpenVINO](https://docs.openvino.ai/latest/index.html) | `openvino` | `yolov8n-pose_openvino_model/` | ✅ | `imgsz`, `half` |
| [OpenVINO](../integrations/openvino.md) | `openvino` | `yolov8n-pose_openvino_model/` | ✅ | `imgsz`, `half`, `int8` |
| [TensorRT](https://developer.nvidia.com/tensorrt) | `engine` | `yolov8n-pose.engine` | ✅ | `imgsz`, `half`, `dynamic`, `simplify`, `workspace` |
| [CoreML](https://github.com/apple/coremltools) | `coreml` | `yolov8n-pose.mlpackage` | ✅ | `imgsz`, `half`, `int8`, `nms` |
| [TF SavedModel](https://www.tensorflow.org/guide/saved_model) | `saved_model` | `yolov8n-pose_saved_model/` | ✅ | `imgsz`, `keras` |

2
docs/en/tasks/segment.md
Unescape View File

@ -173,7 +173,7 @@ Available YOLOv8-seg export formats are in the table below. You can predict or v
| [PyTorch](https://pytorch.org/) | - | `yolov8n-seg.pt` | ✅ | - |
| [TorchScript](https://pytorch.org/docs/stable/jit.html) | `torchscript` | `yolov8n-seg.torchscript` | ✅ | `imgsz`, `optimize` |
| [ONNX](https://onnx.ai/) | `onnx` | `yolov8n-seg.onnx` | ✅ | `imgsz`, `half`, `dynamic`, `simplify`, `opset` |
| [OpenVINO](https://docs.openvino.ai/latest/index.html) | `openvino` | `yolov8n-seg_openvino_model/` | ✅ | `imgsz`, `half` |
| [OpenVINO](../integrations/openvino.md) | `openvino` | `yolov8n-seg_openvino_model/` | ✅ | `imgsz`, `half`, `int8` |
| [TensorRT](https://developer.nvidia.com/tensorrt) | `engine` | `yolov8n-seg.engine` | ✅ | `imgsz`, `half`, `dynamic`, `simplify`, `workspace` |
| [CoreML](https://github.com/apple/coremltools) | `coreml` | `yolov8n-seg.mlpackage` | ✅ | `imgsz`, `half`, `int8`, `nms` |
| [TF SavedModel](https://www.tensorflow.org/guide/saved_model) | `saved_model` | `yolov8n-seg_saved_model/` | ✅ | `imgsz`, `keras` |

2
docs/en/usage/cli.md
Unescape View File

@ -175,7 +175,7 @@ Available YOLOv8 export formats are in the table below. You can export to any fo
| [PyTorch](https://pytorch.org/) | - | `yolov8n.pt` | ✅ | - |
| [TorchScript](https://pytorch.org/docs/stable/jit.html) | `torchscript` | `yolov8n.torchscript` | ✅ | `imgsz`, `optimize` |
| [ONNX](https://onnx.ai/) | `onnx` | `yolov8n.onnx` | ✅ | `imgsz`, `half`, `dynamic`, `simplify`, `opset` |
| [OpenVINO](https://docs.openvino.ai/latest/index.html) | `openvino` | `yolov8n_openvino_model/` | ✅ | `imgsz`, `half`, `int8` |
| [OpenVINO](../integrations/openvino.md) | `openvino` | `yolov8n_openvino_model/` | ✅ | `imgsz`, `half`, `int8` |
| [TensorRT](https://developer.nvidia.com/tensorrt) | `engine` | `yolov8n.engine` | ✅ | `imgsz`, `half`, `dynamic`, `simplify`, `workspace` |
| [CoreML](https://github.com/apple/coremltools) | `coreml` | `yolov8n.mlpackage` | ✅ | `imgsz`, `half`, `int8`, `nms` |
| [TF SavedModel](https://www.tensorflow.org/guide/saved_model) | `saved_model` | `yolov8n_saved_model/` | ✅ | `imgsz`, `keras`, `int8` |

1
docs/en/usage/python.md
Unescape View File

@ -286,7 +286,6 @@ Explorer API can be used to explore datasets with advanced semantic, vector-simi
[Explorer](../datasets/explorer/index.md){ .md-button }
## Using Trainers
`YOLO` model class is a high-level wrapper on the Trainer classes. Each YOLO task has its own trainer that inherits from `BaseTrainer`.

20
docs/en/yolov5/tutorials/model_export.md
Unescape View File

@ -182,16 +182,16 @@ import torch
# Model
model = torch.hub.load('ultralytics/yolov5', 'custom', 'yolov5s.pt')
'yolov5s.torchscript ') # TorchScript
'yolov5s.onnx') # ONNX Runtime
'yolov5s_openvino_model') # OpenVINO
'yolov5s.engine') # TensorRT
'yolov5s.mlmodel') # CoreML (macOS Only)
'yolov5s_saved_model') # TensorFlow SavedModel
'yolov5s.pb') # TensorFlow GraphDef
'yolov5s.tflite') # TensorFlow Lite
'yolov5s_edgetpu.tflite') # TensorFlow Edge TPU
'yolov5s_paddle_model') # PaddlePaddle
model = torch.hub.load('ultralytics/yolov5', 'custom', 'yolov5s.torchscript ') # TorchScript
model = torch.hub.load('ultralytics/yolov5', 'custom', 'yolov5s.onnx') # ONNX Runtime
model = torch.hub.load('ultralytics/yolov5', 'custom', 'yolov5s_openvino_model') # OpenVINO
model = torch.hub.load('ultralytics/yolov5', 'custom', 'yolov5s.engine') # TensorRT
model = torch.hub.load('ultralytics/yolov5', 'custom', 'yolov5s.mlmodel') # CoreML (macOS Only)
model = torch.hub.load('ultralytics/yolov5', 'custom', 'yolov5s_saved_model') # TensorFlow SavedModel
model = torch.hub.load('ultralytics/yolov5', 'custom', 'yolov5s.pb') # TensorFlow GraphDef
model = torch.hub.load('ultralytics/yolov5', 'custom', 'yolov5s.tflite') # TensorFlow Lite
model = torch.hub.load('ultralytics/yolov5', 'custom', 'yolov5s_edgetpu.tflite') # TensorFlow Edge TPU
model = torch.hub.load('ultralytics/yolov5', 'custom', 'yolov5s_paddle_model') # PaddlePaddle
# Images
img = 'https://ultralytics.com/images/zidane.jpg' # or file, Path, PIL, OpenCV, numpy, list

13
ultralytics/engine/exporter.py
Unescape View File

@ -429,12 +429,20 @@ class Exporter:
) # export
if self.args.int8:
assert self.args.data, "INT8 export requires a data argument for calibration, i.e. 'data=coco8.yaml'"
if not self.args.data:
self.args.data = DEFAULT_CFG.data or "coco128.yaml"
LOGGER.warning(
f"{prefix} WARNING ⚠️ INT8 export requires a missing 'data' arg for calibration. "
f"Using default 'data={self.args.data}'."
)
check_requirements("nncf>=2.5.0")
import nncf
def transform_fn(data_item):
"""Quantization transform function."""
assert (
data_item["img"].dtype == torch.uint8
), "Input image must be uint8 for the quantization preprocessing"
im = data_item["img"].numpy().astype(np.float32) / 255.0 # uint8 to fp16/32 and 0 - 255 to 0.0 - 1.0
return np.expand_dims(im, 0) if im.ndim == 3 else im
@ -442,6 +450,9 @@ class Exporter:
LOGGER.info(f"{prefix} collecting INT8 calibration images from 'data={self.args.data}'")
data = check_det_dataset(self.args.data)
dataset = YOLODataset(data["val"], data=data, imgsz=self.imgsz[0], augment=False)
n = len(dataset)
if n < 300:
LOGGER.warning(f"{prefix} WARNING ⚠️ >300 images recommended for INT8 calibration, found {n} images.")
quantization_dataset = nncf.Dataset(dataset, transform_fn)
ignored_scope = nncf.IgnoredScope(types=["Multiply", "Subtract", "Sigmoid"]) # ignore operation
quantized_ov_model = nncf.quantize(

Write Preview
Loading…
Cancel
Save