Spaces:
Sleeping
Sleeping
Added yolov8 model
#2
by
kadirnar
- opened
- README.md +1 -1
- app.py +74 -83
- data/26.jpg +3 -0
- data/27.jpg +3 -0
- data/28.jpg +3 -0
- data/31.jpg +3 -0
- requirements.txt +2 -3
README.md
CHANGED
@@ -6,7 +6,7 @@ colorTo: yellow
|
|
6 |
sdk: gradio
|
7 |
app_file: app.py
|
8 |
pinned: false
|
9 |
-
duplicated_from:
|
10 |
license: openrail
|
11 |
---
|
12 |
|
|
|
6 |
sdk: gradio
|
7 |
app_file: app.py
|
8 |
pinned: false
|
9 |
+
duplicated_from: deprem-ml/deprem_satellite_test
|
10 |
license: openrail
|
11 |
---
|
12 |
|
app.py
CHANGED
@@ -1,34 +1,21 @@
|
|
1 |
-
import
|
2 |
-
import sahi.utils
|
3 |
-
from sahi import AutoDetectionModel
|
4 |
-
import sahi.predict
|
5 |
-
import sahi.slicing
|
6 |
from PIL import Image
|
|
|
7 |
import numpy
|
8 |
-
from huggingface_hub import hf_hub_download
|
9 |
import torch
|
|
|
10 |
|
|
|
11 |
|
12 |
-
|
13 |
-
|
14 |
-
model_id = 'deprem-ml/Binafarktespit-yolo5x-v1-xview'
|
15 |
-
|
16 |
-
|
17 |
current_device = "cuda" if torch.cuda.is_available() else "cpu"
|
18 |
-
model_types = ["YOLOv5", "YOLOv5 + SAHI"]
|
19 |
-
# Model
|
20 |
-
model = AutoDetectionModel.from_pretrained(
|
21 |
-
model_type="yolov5",
|
22 |
-
model_path=model_id,
|
23 |
-
device=current_device,
|
24 |
-
confidence_threshold=0.5,
|
25 |
-
image_size=IMAGE_SIZE,
|
26 |
-
)
|
27 |
-
|
28 |
|
29 |
-
def
|
30 |
-
model_type,
|
31 |
image,
|
|
|
|
|
|
|
32 |
slice_height=512,
|
33 |
slice_width=512,
|
34 |
overlap_height_ratio=0.1,
|
@@ -39,26 +26,43 @@ def sahi_yolo_inference(
|
|
39 |
postprocess_class_agnostic=False,
|
40 |
):
|
41 |
|
42 |
-
# image_width, image_height = image.size
|
43 |
-
# sliced_bboxes = sahi.slicing.get_slice_bboxes(
|
44 |
-
# image_height,
|
45 |
-
# image_width,
|
46 |
-
# slice_height,
|
47 |
-
# slice_width,
|
48 |
-
# False,
|
49 |
-
# overlap_height_ratio,
|
50 |
-
# overlap_width_ratio,
|
51 |
-
# )
|
52 |
-
# if len(sliced_bboxes) > 60:
|
53 |
-
# raise ValueError(
|
54 |
-
# f"{len(sliced_bboxes)} slices are too much for huggingface spaces, try smaller slice size."
|
55 |
-
# )
|
56 |
-
|
57 |
rect_th = None or max(round(sum(image.size) / 2 * 0.0001), 1)
|
58 |
text_th = None or max(rect_th - 2, 1)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
59 |
|
60 |
-
|
61 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
62 |
image=image,
|
63 |
detection_model=model,
|
64 |
slice_height=int(slice_height),
|
@@ -70,54 +74,38 @@ def sahi_yolo_inference(
|
|
70 |
postprocess_match_threshold=postprocess_match_threshold,
|
71 |
postprocess_class_agnostic=postprocess_class_agnostic,
|
72 |
)
|
73 |
-
|
|
|
74 |
image=numpy.array(image),
|
75 |
object_prediction_list=prediction_result_2.object_prediction_list,
|
76 |
rect_th=rect_th,
|
77 |
text_th=text_th,
|
78 |
)
|
|
|
79 |
output = Image.fromarray(visual_result_2["image"])
|
80 |
return output
|
81 |
|
82 |
-
|
83 |
-
|
84 |
-
prediction_result_1 = sahi.predict.get_prediction(
|
85 |
-
image=image, detection_model=model
|
86 |
-
)
|
87 |
-
print(image)
|
88 |
-
visual_result_1 = sahi.utils.cv.visualize_object_predictions(
|
89 |
-
image=numpy.array(image),
|
90 |
-
object_prediction_list=prediction_result_1.object_prediction_list,
|
91 |
-
rect_th=rect_th,
|
92 |
-
text_th=text_th,
|
93 |
-
)
|
94 |
-
output = Image.fromarray(visual_result_1["image"])
|
95 |
-
return output
|
96 |
-
|
97 |
-
# sliced inference
|
98 |
|
|
|
|
|
|
|
|
|
99 |
|
100 |
inputs = [
|
101 |
-
gr.Dropdown(
|
102 |
-
choices=model_types,
|
103 |
-
label="Choose Model Type",
|
104 |
-
type="value",
|
105 |
-
value=model_types[1],
|
106 |
-
),
|
107 |
gr.Image(type="pil", label="Original Image"),
|
108 |
-
gr.
|
109 |
-
gr.
|
110 |
-
gr.
|
111 |
-
gr.
|
112 |
-
gr.
|
113 |
-
|
114 |
-
|
115 |
-
|
116 |
-
|
117 |
-
),
|
118 |
-
gr.
|
119 |
-
gr.Number(value=0.5, label="postprocess_match_threshold"),
|
120 |
-
gr.Checkbox(value=True, label="postprocess_class_agnostic"),
|
121 |
]
|
122 |
|
123 |
outputs = [gr.outputs.Image(type="pil", label="Output")]
|
@@ -126,13 +114,14 @@ title = "Building Detection from Satellite Images with SAHI + YOLOv5"
|
|
126 |
description = "SAHI + YOLOv5 demo for building detection from satellite images. Upload an image or click an example image to use."
|
127 |
article = "<p style='text-align: center'>SAHI is a lightweight vision library for performing large scale object detection/ instance segmentation.. <a href='https://github.com/obss/sahi'>SAHI Github</a> | <a href='https://medium.com/codable/sahi-a-vision-library-for-performing-sliced-inference-on-large-images-small-objects-c8b086af3b80'>SAHI Blog</a> | <a href='https://github.com/fcakyon/yolov5-pip'>YOLOv5 Github</a> </p>"
|
128 |
examples = [
|
129 |
-
[
|
130 |
-
[
|
131 |
-
[
|
132 |
-
[
|
133 |
]
|
134 |
-
|
135 |
-
|
|
|
136 |
inputs,
|
137 |
outputs,
|
138 |
title=title,
|
@@ -141,4 +130,6 @@ gr.Interface(
|
|
141 |
examples=examples,
|
142 |
theme="huggingface",
|
143 |
cache_examples=True,
|
144 |
-
)
|
|
|
|
|
|
1 |
+
from sahi import utils, predict, AutoDetectionModel
|
|
|
|
|
|
|
|
|
2 |
from PIL import Image
|
3 |
+
import gradio as gr
|
4 |
import numpy
|
|
|
5 |
import torch
|
6 |
+
import os
|
7 |
|
8 |
+
os.system('pip install git+https://github.com/fcakyon/ultralyticsplus.git')
|
9 |
|
10 |
+
model_id_list = ['deprem-ml/Binafarktespit-yolo5x-v1-xview', 'SerdarHelli/deprem_satellite_labeled_yolov8']
|
|
|
|
|
|
|
|
|
11 |
current_device = "cuda" if torch.cuda.is_available() else "cpu"
|
12 |
+
model_types = ["YOLOv5", "YOLOv5 + SAHI", "YOLOv8"]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
13 |
|
14 |
+
def sahi_yolov5_inference(
|
|
|
15 |
image,
|
16 |
+
model_id,
|
17 |
+
model_type,
|
18 |
+
image_size,
|
19 |
slice_height=512,
|
20 |
slice_width=512,
|
21 |
overlap_height_ratio=0.1,
|
|
|
26 |
postprocess_class_agnostic=False,
|
27 |
):
|
28 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
29 |
rect_th = None or max(round(sum(image.size) / 2 * 0.0001), 1)
|
30 |
text_th = None or max(rect_th - 2, 1)
|
31 |
+
|
32 |
+
if model_type == "YOLOv5":
|
33 |
+
# standard inference
|
34 |
+
model = AutoDetectionModel.from_pretrained(
|
35 |
+
model_type="yolov5",
|
36 |
+
model_path=model_id,
|
37 |
+
device=current_device,
|
38 |
+
confidence_threshold=0.5,
|
39 |
+
image_size=image_size,
|
40 |
+
)
|
41 |
+
|
42 |
+
prediction_result_1 = predict.get_prediction(
|
43 |
+
image=image, detection_model=model
|
44 |
+
)
|
45 |
|
46 |
+
visual_result_1 = utils.cv.visualize_object_predictions(
|
47 |
+
image=numpy.array(image),
|
48 |
+
object_prediction_list=prediction_result_1.object_prediction_list,
|
49 |
+
rect_th=rect_th,
|
50 |
+
text_th=text_th,
|
51 |
+
)
|
52 |
+
|
53 |
+
output = Image.fromarray(visual_result_1["image"])
|
54 |
+
return output
|
55 |
+
|
56 |
+
elif model_type == "YOLOv5 + SAHI":
|
57 |
+
model = AutoDetectionModel.from_pretrained(
|
58 |
+
model_type="yolov5",
|
59 |
+
model_path=model_id,
|
60 |
+
device=current_device,
|
61 |
+
confidence_threshold=0.5,
|
62 |
+
image_size=image_size,
|
63 |
+
)
|
64 |
+
|
65 |
+
prediction_result_2 = predict.get_sliced_prediction(
|
66 |
image=image,
|
67 |
detection_model=model,
|
68 |
slice_height=int(slice_height),
|
|
|
74 |
postprocess_match_threshold=postprocess_match_threshold,
|
75 |
postprocess_class_agnostic=postprocess_class_agnostic,
|
76 |
)
|
77 |
+
|
78 |
+
visual_result_2 = utils.cv.visualize_object_predictions(
|
79 |
image=numpy.array(image),
|
80 |
object_prediction_list=prediction_result_2.object_prediction_list,
|
81 |
rect_th=rect_th,
|
82 |
text_th=text_th,
|
83 |
)
|
84 |
+
|
85 |
output = Image.fromarray(visual_result_2["image"])
|
86 |
return output
|
87 |
|
88 |
+
elif model_type == "YOLOv8":
|
89 |
+
from ultralyticsplus import YOLO, render_result
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
90 |
|
91 |
+
model = YOLO('SerdarHelli/deprem_satellite_labeled_yolov8')
|
92 |
+
result = model.predict(image, imgsz=image_size)[0]
|
93 |
+
render = render_result(model=model, image=image, result=result, rect_th=rect_th, text_th=text_th)
|
94 |
+
return render
|
95 |
|
96 |
inputs = [
|
|
|
|
|
|
|
|
|
|
|
|
|
97 |
gr.Image(type="pil", label="Original Image"),
|
98 |
+
gr.Dropdown(choices=model_id_list,label="Choose Model",value=model_id_list[0]),
|
99 |
+
gr.Dropdown( choices=model_types, label="Choose Model Type", value=model_types[1]),
|
100 |
+
gr.Slider(minimum=128, maximum=2048, value=640, step=32, label="Image Size"),
|
101 |
+
gr.Slider(minimum=128, maximum=2048, value=512, step=32, label="Slice Height"),
|
102 |
+
gr.Slider(minimum=128, maximum=2048, value=512, step=32, label="Slice Width"),
|
103 |
+
gr.Slider(minimum=0.0, maximum=1.0, value=0.1, step=0.1, label="Overlap Height Ratio"),
|
104 |
+
gr.Slider(minimum=0.0, maximum=1.0, value=0.1, step=0.1, label="Overlap Width Ratio"),
|
105 |
+
gr.Dropdown(["NMS", "GREEDYNMM"], type="value", value="NMS", label="Postprocess Type"),
|
106 |
+
gr.Dropdown(["IOU", "IOS"], type="value", value="IOU", label="Postprocess Type"),
|
107 |
+
gr.Slider(minimum=0.0, maximum=1.0, value=0.5, step=0.1, label="Postprocess Match Threshold"),
|
108 |
+
gr.Checkbox(value=True, label="Postprocess Class Agnostic"),
|
|
|
|
|
109 |
]
|
110 |
|
111 |
outputs = [gr.outputs.Image(type="pil", label="Output")]
|
|
|
114 |
description = "SAHI + YOLOv5 demo for building detection from satellite images. Upload an image or click an example image to use."
|
115 |
article = "<p style='text-align: center'>SAHI is a lightweight vision library for performing large scale object detection/ instance segmentation.. <a href='https://github.com/obss/sahi'>SAHI Github</a> | <a href='https://medium.com/codable/sahi-a-vision-library-for-performing-sliced-inference-on-large-images-small-objects-c8b086af3b80'>SAHI Blog</a> | <a href='https://github.com/fcakyon/yolov5-pip'>YOLOv5 Github</a> </p>"
|
116 |
examples = [
|
117 |
+
["data/26.jpg", 'deprem-ml/Binafarktespit-yolo5x-v1-xview', "YOLOv5 + SAHI", 640, 512, 512, 0.1, 0.1, "NMS", "IOU", 0.25, False],
|
118 |
+
["data/27.jpg", 'deprem-ml/Binafarktespit-yolo5x-v1-xview', "YOLOv5 + SAHI", 640, 512, 512, 0.1, 0.1, "NMS", "IOU", 0.25, False],
|
119 |
+
["data/28.jpg", 'deprem-ml/Binafarktespit-yolo5x-v1-xview', "YOLOv5 + SAHI", 640, 512, 512, 0.1, 0.1, "NMS", "IOU", 0.25, False],
|
120 |
+
["data/31.jpg", 'deprem-ml/SerdarHelli-yolov8-v1-xview', "YOLOv8", 640, 512, 512, 0.1, 0.1, "NMS", "IOU", 0.25, False],
|
121 |
]
|
122 |
+
|
123 |
+
demo = gr.Interface(
|
124 |
+
sahi_yolov5_inference,
|
125 |
inputs,
|
126 |
outputs,
|
127 |
title=title,
|
|
|
130 |
examples=examples,
|
131 |
theme="huggingface",
|
132 |
cache_examples=True,
|
133 |
+
)
|
134 |
+
|
135 |
+
demo.launch(debug=True, enable_queue=True)
|
data/26.jpg
ADDED
![]() |
Git LFS Details
|
data/27.jpg
ADDED
![]() |
Git LFS Details
|
data/28.jpg
ADDED
![]() |
Git LFS Details
|
data/31.jpg
ADDED
![]() |
Git LFS Details
|
requirements.txt
CHANGED
@@ -1,5 +1,4 @@
|
|
1 |
-
torch==1.10.2
|
2 |
-
torchvision==0.11.3
|
3 |
-
-f https://download.pytorch.org/whl/torch_stable.html
|
4 |
yolov5==7.0.8
|
5 |
sahi==0.11.11
|
|
|
1 |
+
torch==1.10.2
|
2 |
+
torchvision==0.11.3
|
|
|
3 |
yolov5==7.0.8
|
4 |
sahi==0.11.11
|