SAHI with YOLOv5 for Sliced Inference¶

0. Preparation¶

• Install latest version of SAHI and YOLOv5:

In [ ]:

!pip install -U torch sahi==0.11.21 yolov5==7.0.14

!pip install -U torch sahi==0.11.21 yolov5==7.0.14

In [ ]:

import os
os.getcwd()

import os os.getcwd()

• Import required modules:

In [1]:

# arrange an instance segmentation model for test
from sahi.utils.yolov5 import (
    download_yolov5s6_model,
)

# import required functions, classes
from sahi import AutoDetectionModel
from sahi.utils.cv import read_image
from sahi.utils.file import download_from_url
from sahi.predict import get_prediction, get_sliced_prediction, predict
from IPython.display import Image

# arrange an instance segmentation model for test from sahi.utils.yolov5 import ( download_yolov5s6_model, ) # import required functions, classes from sahi import AutoDetectionModel from sahi.utils.cv import read_image from sahi.utils.file import download_from_url from sahi.predict import get_prediction, get_sliced_prediction, predict from IPython.display import Image

• Download a yolov5 model and two test images:

In [2]:

# download YOLOV5S6 model to 'models/yolov5s6.pt'
yolov5_model_path = 'models/yolov5s6.pt'
download_yolov5s6_model(destination_path=yolov5_model_path)

# download test images into demo_data folder
download_from_url('https://raw.githubusercontent.com/obss/sahi/main/demo/demo_data/small-vehicles1.jpeg', 'demo_data/small-vehicles1.jpeg')
download_from_url('https://raw.githubusercontent.com/obss/sahi/main/demo/demo_data/terrain2.png', 'demo_data/terrain2.png')

# download YOLOV5S6 model to 'models/yolov5s6.pt' yolov5_model_path = 'models/yolov5s6.pt' download_yolov5s6_model(destination_path=yolov5_model_path) # download test images into demo_data folder download_from_url('https://raw.githubusercontent.com/obss/sahi/main/demo/demo_data/small-vehicles1.jpeg', 'demo_data/small-vehicles1.jpeg') download_from_url('https://raw.githubusercontent.com/obss/sahi/main/demo/demo_data/terrain2.png', 'demo_data/terrain2.png')

1. Standard Inference with a YOLOv5 Model¶

• Instantiate a detection model by defining model weight path and other parameters:

In [18]:

detection_model = AutoDetectionModel.from_pretrained(
    model_type='yolov5',
    model_path=yolov5_model_path,
    confidence_threshold=0.3,
    device="cpu", # or 'cuda:0'
)

detection_model = AutoDetectionModel.from_pretrained( model_type='yolov5', model_path=yolov5_model_path, confidence_threshold=0.3, device="cpu", # or 'cuda:0' )

• Perform prediction by feeding the get_prediction function with an image path and a DetectionModel instance:

In [4]:

result = get_prediction("demo_data/small-vehicles1.jpeg", detection_model)

result = get_prediction("demo_data/small-vehicles1.jpeg", detection_model)

• Or perform prediction by feeding the get_prediction function with a numpy image and a DetectionModel instance:

In [5]:

result = get_prediction(read_image("demo_data/small-vehicles1.jpeg"), detection_model)

result = get_prediction(read_image("demo_data/small-vehicles1.jpeg"), detection_model)

• Visualize predicted bounding boxes and masks over the original image:

In [6]:

result.export_visuals(export_dir="demo_data/")

Image("demo_data/prediction_visual.png")

result.export_visuals(export_dir="demo_data/") Image("demo_data/prediction_visual.png")

Out[6]:

2. Sliced Inference with a YOLOv5 Model¶

• To perform sliced prediction we need to specify slice parameters. In this example we will perform prediction over slices of 256x256 with an overlap ratio of 0.2:

In [7]:

result = get_sliced_prediction(
    "demo_data/small-vehicles1.jpeg",
    detection_model,
    slice_height = 256,
    slice_width = 256,
    overlap_height_ratio = 0.2,
    overlap_width_ratio = 0.2
)

result = get_sliced_prediction( "demo_data/small-vehicles1.jpeg", detection_model, slice_height = 256, slice_width = 256, overlap_height_ratio = 0.2, overlap_width_ratio = 0.2 )

Performing prediction on 15 number of slices.

• Visualize predicted bounding boxes and masks over the original image:

In [8]:

result.export_visuals(export_dir="demo_data/")

Image("demo_data/prediction_visual.png")

result.export_visuals(export_dir="demo_data/") Image("demo_data/prediction_visual.png")

Out[8]:

3. Prediction Result¶

• Predictions are returned as sahi.prediction.PredictionResult, you can access the object prediction list as:

In [9]:

object_prediction_list = result.object_prediction_list

object_prediction_list = result.object_prediction_list

In [10]:

object_prediction_list[0]

object_prediction_list[0]

Out[10]:

ObjectPrediction<
    bbox: BoundingBox: <(447, 308, 496, 342), w: 49, h: 34>,
    mask: None,
    score: PredictionScore: <value: 0.9154329299926758>,
    category: Category: <id: 2, name: car>>

• ObjectPrediction's can be converted to COCO annotation format:

In [11]:

result.to_coco_annotations()[:3]

result.to_coco_annotations()[:3]

Out[11]:

[{'image_id': None,
  'bbox': [447, 308, 49, 34],
  'score': 0.9154329299926758,
  'category_id': 2,
  'category_name': 'car',
  'segmentation': [],
  'iscrowd': 0,
  'area': 1666},
 {'image_id': None,
  'bbox': [321, 321, 62, 41],
  'score': 0.887977659702301,
  'category_id': 2,
  'category_name': 'car',
  'segmentation': [],
  'iscrowd': 0,
  'area': 2542},
 {'image_id': None,
  'bbox': [382, 278, 37, 26],
  'score': 0.8796820640563965,
  'category_id': 2,
  'category_name': 'car',
  'segmentation': [],
  'iscrowd': 0,
  'area': 962}]

• ObjectPrediction's can be converted to COCO prediction format:

In [12]:

result.to_coco_predictions(image_id=1)[:3]

result.to_coco_predictions(image_id=1)[:3]

Out[12]:

[{'image_id': 1,
  'bbox': [447, 308, 49, 34],
  'score': 0.9154329299926758,
  'category_id': 2,
  'category_name': 'car',
  'segmentation': [],
  'iscrowd': 0,
  'area': 1666},
 {'image_id': 1,
  'bbox': [321, 321, 62, 41],
  'score': 0.887977659702301,
  'category_id': 2,
  'category_name': 'car',
  'segmentation': [],
  'iscrowd': 0,
  'area': 2542},
 {'image_id': 1,
  'bbox': [382, 278, 37, 26],
  'score': 0.8796820640563965,
  'category_id': 2,
  'category_name': 'car',
  'segmentation': [],
  'iscrowd': 0,
  'area': 962}]

• ObjectPrediction's can be converted to imantics annotation format:

In [13]:

result.to_imantics_annotations()[:3]

result.to_imantics_annotations()[:3]

Out[13]:

[<imantics.annotation.Annotation at 0x7f9d84c8fa60>,
 <imantics.annotation.Annotation at 0x7f9d84d118e0>,
 <imantics.annotation.Annotation at 0x7f9d84d11e50>]

• ObjectPrediction's can be converted to fiftyone detection format:

In [14]:

result.to_fiftyone_detections()[:3]

result.to_fiftyone_detections()[:3]

Out[14]:

[<Detection: {
     'id': '633305c1e757fa81eb2b8348',
     'attributes': BaseDict({}),
     'tags': BaseList([]),
     'label': 'car',
     'bounding_box': BaseList([
         0.41853932584269665,
         0.5310344827586206,
         0.04588014981273408,
         0.05862068965517241,
     ]),
     'mask': None,
     'confidence': 0.9154329299926758,
     'index': None,
 }>,
 <Detection: {
     'id': '633305c1e757fa81eb2b8349',
     'attributes': BaseDict({}),
     'tags': BaseList([]),
     'label': 'car',
     'bounding_box': BaseList([
         0.300561797752809,
         0.553448275862069,
         0.05805243445692884,
         0.0706896551724138,
     ]),
     'mask': None,
     'confidence': 0.887977659702301,
     'index': None,
 }>,
 <Detection: {
     'id': '633305c1e757fa81eb2b834a',
     'attributes': BaseDict({}),
     'tags': BaseList([]),
     'label': 'car',
     'bounding_box': BaseList([
         0.35767790262172283,
         0.4793103448275862,
         0.03464419475655431,
         0.04482758620689655,
     ]),
     'mask': None,
     'confidence': 0.8796820640563965,
     'index': None,
 }>]

4. Batch Prediction¶

• Set model and directory parameters:

In [3]:

model_type = "yolov5"
model_path = yolov5_model_path
model_device = "cpu" # or 'cuda:0'
model_confidence_threshold = 0.4

slice_height = 256
slice_width = 256
overlap_height_ratio = 0.2
overlap_width_ratio = 0.2

source_image_dir = "demo_data/"

model_type = "yolov5" model_path = yolov5_model_path model_device = "cpu" # or 'cuda:0' model_confidence_threshold = 0.4 slice_height = 256 slice_width = 256 overlap_height_ratio = 0.2 overlap_width_ratio = 0.2 source_image_dir = "demo_data/"

• Perform sliced inference on given folder:

In [4]:

predict(
    model_type=model_type,
    model_path=model_path,
    model_device=model_device,
    model_confidence_threshold=model_confidence_threshold,
    source=source_image_dir,
    slice_height=slice_height,
    slice_width=slice_width,
    overlap_height_ratio=overlap_height_ratio,
    overlap_width_ratio=overlap_width_ratio,
)

predict( model_type=model_type, model_path=model_path, model_device=model_device, model_confidence_threshold=model_confidence_threshold, source=source_image_dir, slice_height=slice_height, slice_width=slice_width, overlap_height_ratio=overlap_height_ratio, overlap_width_ratio=overlap_width_ratio, )

There are 3 listed files in folder: demo_data/

Performing inference on images:   0%|          | 0/3 [00:00<?, ?it/s]

Performing prediction on 20 number of slices.

Performing inference on images:  33%|███▎      | 1/3 [00:01<00:02,  1.01s/it]

Prediction time is: 971.29 ms
Performing prediction on 15 number of slices.

Performing inference on images:  67%|██████▋   | 2/3 [00:01<00:00,  1.59it/s]

Prediction time is: 318.99 ms
Performing prediction on 15 number of slices.

Performing inference on images: 100%|██████████| 3/3 [00:01<00:00,  1.77it/s]

Prediction time is: 292.02 ms
Prediction results are successfully exported to runs/predict/exp12