Spaces:
Running
Running
First commit
Browse files- .gitignore +162 -0
- README.md +6 -6
- app.py +170 -111
.gitignore
ADDED
@@ -0,0 +1,162 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
### Python template
|
2 |
+
# Byte-compiled / optimized / DLL files
|
3 |
+
__pycache__/
|
4 |
+
*.py[cod]
|
5 |
+
*$py.class
|
6 |
+
.idea/
|
7 |
+
# C extensions
|
8 |
+
*.so
|
9 |
+
|
10 |
+
# Distribution / packaging
|
11 |
+
.Python
|
12 |
+
build/
|
13 |
+
develop-eggs/
|
14 |
+
dist/
|
15 |
+
downloads/
|
16 |
+
eggs/
|
17 |
+
.eggs/
|
18 |
+
lib/
|
19 |
+
lib64/
|
20 |
+
parts/
|
21 |
+
sdist/
|
22 |
+
var/
|
23 |
+
wheels/
|
24 |
+
share/python-wheels/
|
25 |
+
*.egg-info/
|
26 |
+
.installed.cfg
|
27 |
+
*.egg
|
28 |
+
MANIFEST
|
29 |
+
|
30 |
+
# PyInstaller
|
31 |
+
# Usually these files are written by a python script from a template
|
32 |
+
# before PyInstaller builds the exe, so as to inject date/other infos into it.
|
33 |
+
*.manifest
|
34 |
+
*.spec
|
35 |
+
|
36 |
+
# Installer logs
|
37 |
+
pip-log.txt
|
38 |
+
pip-delete-this-directory.txt
|
39 |
+
|
40 |
+
# Unit test / coverage reports
|
41 |
+
htmlcov/
|
42 |
+
.tox/
|
43 |
+
.nox/
|
44 |
+
.coverage
|
45 |
+
.coverage.*
|
46 |
+
.cache
|
47 |
+
nosetests.xml
|
48 |
+
coverage.xml
|
49 |
+
*.cover
|
50 |
+
*.py,cover
|
51 |
+
.hypothesis/
|
52 |
+
.pytest_cache/
|
53 |
+
cover/
|
54 |
+
|
55 |
+
# Translations
|
56 |
+
*.mo
|
57 |
+
*.pot
|
58 |
+
|
59 |
+
# Django stuff:
|
60 |
+
*.log
|
61 |
+
local_settings.py
|
62 |
+
db.sqlite3
|
63 |
+
db.sqlite3-journal
|
64 |
+
|
65 |
+
# Flask stuff:
|
66 |
+
instance/
|
67 |
+
.webassets-cache
|
68 |
+
|
69 |
+
# Scrapy stuff:
|
70 |
+
.scrapy
|
71 |
+
|
72 |
+
# Sphinx documentation
|
73 |
+
docs/_build/
|
74 |
+
|
75 |
+
# PyBuilder
|
76 |
+
.pybuilder/
|
77 |
+
target/
|
78 |
+
|
79 |
+
# Jupyter Notebook
|
80 |
+
.ipynb_checkpoints
|
81 |
+
|
82 |
+
# IPython
|
83 |
+
profile_default/
|
84 |
+
ipython_config.py
|
85 |
+
|
86 |
+
# pyenv
|
87 |
+
# For a library or package, you might want to ignore these files since the code is
|
88 |
+
# intended to run in multiple environments; otherwise, check them in:
|
89 |
+
# .python-version
|
90 |
+
|
91 |
+
# pipenv
|
92 |
+
# According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
|
93 |
+
# However, in case of collaboration, if having platform-specific dependencies or dependencies
|
94 |
+
# having no cross-platform support, pipenv may install dependencies that don't work, or not
|
95 |
+
# install all needed dependencies.
|
96 |
+
#Pipfile.lock
|
97 |
+
|
98 |
+
# poetry
|
99 |
+
# Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
|
100 |
+
# This is especially recommended for binary packages to ensure reproducibility, and is more
|
101 |
+
# commonly ignored for libraries.
|
102 |
+
# https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
|
103 |
+
#poetry.lock
|
104 |
+
|
105 |
+
# pdm
|
106 |
+
# Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
|
107 |
+
#pdm.lock
|
108 |
+
# pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
|
109 |
+
# in version control.
|
110 |
+
# https://pdm.fming.dev/#use-with-ide
|
111 |
+
.pdm.toml
|
112 |
+
|
113 |
+
# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
|
114 |
+
__pypackages__/
|
115 |
+
|
116 |
+
# Celery stuff
|
117 |
+
celerybeat-schedule
|
118 |
+
celerybeat.pid
|
119 |
+
|
120 |
+
# SageMath parsed files
|
121 |
+
*.sage.py
|
122 |
+
|
123 |
+
# Environments
|
124 |
+
.env
|
125 |
+
.venv
|
126 |
+
env/
|
127 |
+
venv/
|
128 |
+
ENV/
|
129 |
+
env.bak/
|
130 |
+
venv.bak/
|
131 |
+
|
132 |
+
# Spyder project settings
|
133 |
+
.spyderproject
|
134 |
+
.spyproject
|
135 |
+
|
136 |
+
# Rope project settings
|
137 |
+
.ropeproject
|
138 |
+
|
139 |
+
# mkdocs documentation
|
140 |
+
/site
|
141 |
+
|
142 |
+
# mypy
|
143 |
+
.mypy_cache/
|
144 |
+
.dmypy.json
|
145 |
+
dmypy.json
|
146 |
+
|
147 |
+
# Pyre type checker
|
148 |
+
.pyre/
|
149 |
+
|
150 |
+
# pytype static type analyzer
|
151 |
+
.pytype/
|
152 |
+
|
153 |
+
# Cython debug symbols
|
154 |
+
cython_debug/
|
155 |
+
|
156 |
+
# PyCharm
|
157 |
+
# JetBrains specific template is maintained in a separate JetBrains.gitignore that can
|
158 |
+
# be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
|
159 |
+
# and can be added to the global gitignore or merged into this file. For a more nuclear
|
160 |
+
# option (not recommended) you can uncomment the following to ignore the entire idea folder.
|
161 |
+
#.idea/
|
162 |
+
|
README.md
CHANGED
@@ -1,14 +1,14 @@
|
|
1 |
---
|
2 |
-
title:
|
3 |
-
emoji:
|
4 |
-
colorFrom:
|
5 |
-
colorTo:
|
6 |
sdk: gradio
|
7 |
-
sdk_version: 5.
|
8 |
app_file: app.py
|
9 |
pinned: false
|
10 |
license: apache-2.0
|
11 |
-
short_description:
|
12 |
---
|
13 |
|
14 |
Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
|
|
|
1 |
---
|
2 |
+
title: Human Interaction Demo
|
3 |
+
emoji: 📊
|
4 |
+
colorFrom: gray
|
5 |
+
colorTo: blue
|
6 |
sdk: gradio
|
7 |
+
sdk_version: 5.6.0
|
8 |
app_file: app.py
|
9 |
pinned: false
|
10 |
license: apache-2.0
|
11 |
+
short_description: Uses pose estimation to determine what are you touching.
|
12 |
---
|
13 |
|
14 |
Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
|
app.py
CHANGED
@@ -1,50 +1,52 @@
|
|
|
|
1 |
import gradio as gr
|
2 |
-
import torch
|
3 |
import numpy as np
|
4 |
-
import cv2
|
5 |
-
from PIL import Image
|
6 |
import supervision as sv
|
|
|
|
|
7 |
from transformers import (
|
8 |
RTDetrForObjectDetection,
|
9 |
RTDetrImageProcessor,
|
10 |
-
VitPoseConfig,
|
11 |
VitPoseForPoseEstimation,
|
12 |
VitPoseImageProcessor,
|
|
|
13 |
)
|
14 |
|
15 |
-
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
class KeypointDetector:
|
38 |
def __init__(self):
|
39 |
self.person_detector = None
|
40 |
self.person_processor = None
|
41 |
self.pose_model = None
|
42 |
self.pose_processor = None
|
|
|
|
|
|
|
43 |
self.load_models()
|
44 |
|
45 |
def load_models(self):
|
46 |
"""Load all required models"""
|
47 |
-
#
|
48 |
self.person_processor = RTDetrImageProcessor.from_pretrained("PekingU/rtdetr_r50vd_coco_o365")
|
49 |
self.person_detector = RTDetrForObjectDetection.from_pretrained("PekingU/rtdetr_r50vd_coco_o365")
|
50 |
|
@@ -52,21 +54,35 @@ class KeypointDetector:
|
|
52 |
self.pose_processor = VitPoseImageProcessor.from_pretrained("nielsr/vitpose-base-simple")
|
53 |
self.pose_model = VitPoseForPoseEstimation.from_pretrained("nielsr/vitpose-base-simple")
|
54 |
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
|
59 |
-
|
60 |
-
|
61 |
-
|
62 |
-
|
63 |
-
|
64 |
-
|
65 |
-
|
66 |
-
|
67 |
-
|
68 |
-
|
69 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
70 |
|
71 |
def detect_persons(self, image: Image.Image):
|
72 |
"""Detect persons in the image"""
|
@@ -80,70 +96,105 @@ class KeypointDetector:
|
|
80 |
threshold=0.3
|
81 |
)
|
82 |
|
83 |
-
# Get boxes and scores for human class (index 0 in COCO dataset)
|
84 |
boxes = results[0]["boxes"][results[0]["labels"] == 0]
|
85 |
scores = results[0]["scores"][results[0]["labels"] == 0]
|
86 |
return boxes.cpu().numpy(), scores.cpu().numpy()
|
87 |
|
88 |
def detect_keypoints(self, image: Image.Image):
|
89 |
"""Detect keypoints in the image"""
|
90 |
-
# Detect persons first
|
91 |
boxes, scores = self.detect_persons(image)
|
92 |
-
boxes_coco = [self.pascal_voc_to_coco(boxes)]
|
93 |
|
94 |
-
|
95 |
-
pixel_values = self.pose_processor(image, boxes=boxes_coco, return_tensors="pt").pixel_values
|
96 |
with torch.no_grad():
|
97 |
outputs = self.pose_model(pixel_values)
|
98 |
|
99 |
-
pose_results = self.pose_processor.post_process_pose_estimation(outputs, boxes=
|
100 |
return pose_results, boxes, scores
|
101 |
|
102 |
-
def
|
103 |
-
"""
|
104 |
-
|
105 |
-
|
|
|
106 |
|
107 |
-
|
108 |
-
|
109 |
-
|
110 |
-
|
111 |
-
|
112 |
-
)
|
|
|
|
|
113 |
|
114 |
-
|
115 |
-
box_annotator = sv.BoxAnnotator(
|
116 |
-
color=sv.ColorPalette.DEFAULT,
|
117 |
-
thickness=2
|
118 |
-
)
|
119 |
|
120 |
-
|
121 |
-
|
122 |
-
|
123 |
-
|
124 |
-
)
|
|
|
|
|
|
|
125 |
|
126 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
127 |
key_points = sv.KeyPoints(
|
128 |
xy=torch.cat([pose_result['keypoints'].unsqueeze(0) for pose_result in pose_results]).cpu().numpy()
|
129 |
)
|
|
|
130 |
|
131 |
-
#
|
132 |
-
|
133 |
-
|
134 |
-
|
135 |
-
)
|
136 |
|
137 |
-
|
138 |
-
|
139 |
-
|
140 |
-
|
141 |
-
|
142 |
|
143 |
-
|
|
|
|
|
|
|
|
|
|
|
144 |
|
145 |
def process_image(self, input_image):
|
146 |
-
"""Process image and return visualization"""
|
147 |
if input_image is None:
|
148 |
return None, ""
|
149 |
|
@@ -153,69 +204,77 @@ class KeypointDetector:
|
|
153 |
else:
|
154 |
image = input_image
|
155 |
|
156 |
-
|
157 |
-
|
|
|
|
|
158 |
|
159 |
# Visualize results
|
160 |
-
result_image = self.
|
161 |
|
162 |
-
# Create
|
163 |
info_text = []
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
164 |
|
165 |
-
#
|
166 |
-
|
167 |
-
|
168 |
-
info_text.append(f"Bounding Box: x1={box[0]:.1f}, y1={box[1]:.1f}, x2={box[2]:.1f}, y2={box[3]:.1f}")
|
169 |
|
170 |
-
|
171 |
-
|
172 |
-
|
173 |
-
|
174 |
-
info_text.append(f"Keypoint {KEYPOINT_LABEL_MAP[j]}: x={x:.1f}, y={y:.1f}, confidence={confidence:.2f}")
|
175 |
|
176 |
-
return result_image, "\n".join(info_text)
|
177 |
|
178 |
|
179 |
def create_gradio_interface():
|
180 |
"""Create Gradio interface"""
|
181 |
-
detector =
|
182 |
|
183 |
with gr.Blocks() as interface:
|
184 |
-
gr.Markdown("#
|
185 |
-
gr.Markdown("Upload an image to detect people
|
186 |
-
gr.Markdown("1. Detect people in the image (shown as bounding boxes)")
|
187 |
-
gr.Markdown("2. Identify keypoints for each detected person (shown as connected green lines)")
|
188 |
-
gr.Markdown("Huge shoutout to @NielsRogge and @SangbumChoi for this work!")
|
189 |
|
190 |
with gr.Row():
|
191 |
with gr.Column():
|
192 |
input_image = gr.Image(label="Input Image")
|
193 |
-
process_button = gr.Button("Detect
|
194 |
|
195 |
with gr.Column():
|
196 |
output_image = gr.Image(label="Detection Results")
|
197 |
-
|
198 |
-
label="
|
199 |
lines=10,
|
200 |
-
placeholder="
|
201 |
)
|
|
|
|
|
202 |
|
203 |
process_button.click(
|
204 |
fn=detector.process_image,
|
205 |
inputs=input_image,
|
206 |
-
outputs=[output_image,
|
207 |
)
|
208 |
|
209 |
gr.Examples(
|
210 |
examples=[
|
211 |
-
"
|
|
|
212 |
],
|
213 |
inputs=input_image
|
214 |
)
|
215 |
|
216 |
return interface
|
217 |
|
218 |
-
|
219 |
if __name__ == "__main__":
|
220 |
-
interface
|
221 |
-
interface.launch()
|
|
|
1 |
+
import cv2
|
2 |
import gradio as gr
|
|
|
3 |
import numpy as np
|
|
|
|
|
4 |
import supervision as sv
|
5 |
+
import torch
|
6 |
+
from PIL import Image
|
7 |
from transformers import (
|
8 |
RTDetrForObjectDetection,
|
9 |
RTDetrImageProcessor,
|
|
|
10 |
VitPoseForPoseEstimation,
|
11 |
VitPoseImageProcessor,
|
12 |
+
pipeline,
|
13 |
)
|
14 |
|
15 |
+
KEYPOINT_LABEL_MAP = {
|
16 |
+
0: "Nose",
|
17 |
+
1: "L_Eye",
|
18 |
+
2: "R_Eye",
|
19 |
+
3: "L_Ear",
|
20 |
+
4: "R_Ear",
|
21 |
+
5: "L_Shoulder",
|
22 |
+
6: "R_Shoulder",
|
23 |
+
7: "L_Elbow",
|
24 |
+
8: "R_Elbow",
|
25 |
+
9: "L_Wrist",
|
26 |
+
10: "R_Wrist",
|
27 |
+
11: "L_Hip",
|
28 |
+
12: "R_Hip",
|
29 |
+
13: "L_Knee",
|
30 |
+
14: "R_Knee",
|
31 |
+
15: "L_Ankle",
|
32 |
+
16: "R_Ankle",
|
33 |
+
}
|
34 |
+
|
35 |
+
|
36 |
+
class InteractionDetector:
|
|
|
37 |
def __init__(self):
|
38 |
self.person_detector = None
|
39 |
self.person_processor = None
|
40 |
self.pose_model = None
|
41 |
self.pose_processor = None
|
42 |
+
self.depth_model = None
|
43 |
+
self.segmentation_model = None
|
44 |
+
self.interaction_threshold = 2
|
45 |
self.load_models()
|
46 |
|
47 |
def load_models(self):
|
48 |
"""Load all required models"""
|
49 |
+
# Person detection model
|
50 |
self.person_processor = RTDetrImageProcessor.from_pretrained("PekingU/rtdetr_r50vd_coco_o365")
|
51 |
self.person_detector = RTDetrForObjectDetection.from_pretrained("PekingU/rtdetr_r50vd_coco_o365")
|
52 |
|
|
|
54 |
self.pose_processor = VitPoseImageProcessor.from_pretrained("nielsr/vitpose-base-simple")
|
55 |
self.pose_model = VitPoseForPoseEstimation.from_pretrained("nielsr/vitpose-base-simple")
|
56 |
|
57 |
+
# Depth estimation model
|
58 |
+
self.depth_model = pipeline("depth-estimation", model="depth-anything/Depth-Anything-V2-Small-hf")
|
59 |
+
|
60 |
+
# Semantic segmentation model
|
61 |
+
self.segmentation_model = pipeline("image-segmentation", model="facebook/maskformer-swin-base-ade")
|
62 |
+
self.segmentation_id2label = self.segmentation_model.model.config.id2label
|
63 |
+
self.segmentation_label2id = {v: k for k, v in self.segmentation_model.model.config.id2label.items()}
|
64 |
+
|
65 |
+
def get_nearest_pixel_class(self, joint, depth_map, segmentation_map):
|
66 |
+
"""
|
67 |
+
Find the nearest pixel of a specific class to a given joint coordinate
|
68 |
+
Args:
|
69 |
+
joint: (x, y) coordinates of the joint
|
70 |
+
depth_map: Depth map
|
71 |
+
segmentation_map: Semantic segmentation results
|
72 |
+
Returns:
|
73 |
+
tuple: class_name of nearest pixel, distance to that pixel
|
74 |
+
"""
|
75 |
+
PERSON_ID = 12
|
76 |
+
grid_x, grid_y = np.meshgrid(np.arange(depth_map.shape[0]), np.arange(depth_map.shape[1]))
|
77 |
+
dist_x = np.abs(grid_x.T - joint[1])
|
78 |
+
dist_y = np.abs(grid_y.T - joint[0])
|
79 |
+
dist_coord = dist_x + dist_y
|
80 |
+
|
81 |
+
|
82 |
+
depth_dist = np.abs(depth_map - depth_map[joint[1], joint[0]])
|
83 |
+
depth_dist[(segmentation_map == PERSON_ID) | (dist_coord > 50)] = 255
|
84 |
+
min_dist = np.unravel_index(np.argmin(depth_dist), depth_dist.shape)
|
85 |
+
return segmentation_map[min_dist], depth_dist[min_dist]
|
86 |
|
87 |
def detect_persons(self, image: Image.Image):
|
88 |
"""Detect persons in the image"""
|
|
|
96 |
threshold=0.3
|
97 |
)
|
98 |
|
|
|
99 |
boxes = results[0]["boxes"][results[0]["labels"] == 0]
|
100 |
scores = results[0]["scores"][results[0]["labels"] == 0]
|
101 |
return boxes.cpu().numpy(), scores.cpu().numpy()
|
102 |
|
103 |
def detect_keypoints(self, image: Image.Image):
|
104 |
"""Detect keypoints in the image"""
|
|
|
105 |
boxes, scores = self.detect_persons(image)
|
|
|
106 |
|
107 |
+
pixel_values = self.pose_processor(image, boxes=[boxes], return_tensors="pt").pixel_values
|
|
|
108 |
with torch.no_grad():
|
109 |
outputs = self.pose_model(pixel_values)
|
110 |
|
111 |
+
pose_results = self.pose_processor.post_process_pose_estimation(outputs, boxes=[boxes])[0]
|
112 |
return pose_results, boxes, scores
|
113 |
|
114 |
+
def estimate_depth(self, image: Image.Image):
|
115 |
+
"""Estimate depth for the image"""
|
116 |
+
with torch.no_grad():
|
117 |
+
depth_map = np.array(self.depth_model(image)['depth'])
|
118 |
+
return depth_map
|
119 |
|
120 |
+
def segment_image(self, image: Image.Image):
|
121 |
+
"""Perform semantic segmentation on the image"""
|
122 |
+
with torch.no_grad():
|
123 |
+
segmentation_map = self.segmentation_model(image)
|
124 |
+
result = np.zeros(np.array(image).shape[:2], dtype=np.uint8)
|
125 |
+
print("Found", [l['label'] for l in segmentation_map])
|
126 |
+
for cls_item in sorted(segmentation_map, key=lambda l: np.sum(l['mask']), reverse=True):
|
127 |
+
result[np.array(cls_item['mask']) > 0] = self.segmentation_label2id[cls_item['label']]
|
128 |
|
129 |
+
return result
|
|
|
|
|
|
|
|
|
130 |
|
131 |
+
def detect_wall_interaction(self, image: Image.Image):
|
132 |
+
"""Detect if hands are touching walls"""
|
133 |
+
# Get all necessary information
|
134 |
+
pose_results, boxes, scores = self.detect_keypoints(image)
|
135 |
+
depth_map = self.estimate_depth(image)
|
136 |
+
segmentation_map = self.segment_image(image)
|
137 |
+
|
138 |
+
interactions = []
|
139 |
|
140 |
+
for person_idx, pose_result in enumerate(pose_results):
|
141 |
+
# Get hand keypoints
|
142 |
+
right_hand = pose_result["keypoints"][10].numpy().astype(int)
|
143 |
+
left_hand = pose_result["keypoints"][9].numpy().astype(int)
|
144 |
+
|
145 |
+
# Find nearest anything pixels
|
146 |
+
right_cls, r_distance = self.get_nearest_pixel_class(right_hand[:2], depth_map, segmentation_map)
|
147 |
+
left_cls, l_distance = self.get_nearest_pixel_class(left_hand[:2], depth_map, segmentation_map)
|
148 |
+
|
149 |
+
|
150 |
+
# Check for interactions
|
151 |
+
right_touching = r_distance < self.interaction_threshold
|
152 |
+
left_touching = l_distance < self.interaction_threshold
|
153 |
+
|
154 |
+
interactions.append({
|
155 |
+
"person_id": person_idx,
|
156 |
+
"right_hand_touching_object": self.segmentation_id2label[right_cls],
|
157 |
+
"left_hand_touching_object": self.segmentation_id2label[left_cls],
|
158 |
+
"right_hand_touching": right_touching,
|
159 |
+
"left_hand_touching": left_touching,
|
160 |
+
"right_hand_distance": r_distance,
|
161 |
+
"left_hand_distance": l_distance
|
162 |
+
})
|
163 |
+
|
164 |
+
return interactions, pose_results, segmentation_map, depth_map
|
165 |
+
|
166 |
+
def visualize_results(self, image: Image.Image, interactions, pose_results):
|
167 |
+
"""Visualize detection results"""
|
168 |
+
# Create base visualization from original image
|
169 |
+
vis_image = np.array(image).copy()
|
170 |
+
|
171 |
+
# Add pose keypoints
|
172 |
+
edge_annotator = sv.EdgeAnnotator(color=sv.Color.GREEN, thickness=2)
|
173 |
key_points = sv.KeyPoints(
|
174 |
xy=torch.cat([pose_result['keypoints'].unsqueeze(0) for pose_result in pose_results]).cpu().numpy()
|
175 |
)
|
176 |
+
vis_image = edge_annotator.annotate(scene=vis_image, key_points=key_points)
|
177 |
|
178 |
+
# Add interaction indicators
|
179 |
+
for interaction in interactions:
|
180 |
+
person_id = interaction["person_id"]
|
181 |
+
pose_result = pose_results[person_id]
|
|
|
182 |
|
183 |
+
# Draw indicators for touching hands
|
184 |
+
if interaction["right_hand_touching"]:
|
185 |
+
cv2.circle(vis_image,
|
186 |
+
tuple(map(int, pose_result["keypoints"][10][:2])),
|
187 |
+
10, (0, 0, 255), -1)
|
188 |
|
189 |
+
if interaction["left_hand_touching"]:
|
190 |
+
cv2.circle(vis_image,
|
191 |
+
tuple(map(int, pose_result["keypoints"][9][:2])),
|
192 |
+
10, (0, 0, 255), -1)
|
193 |
+
|
194 |
+
return Image.fromarray(vis_image)
|
195 |
|
196 |
def process_image(self, input_image):
|
197 |
+
"""Process image and return visualization with interaction detection"""
|
198 |
if input_image is None:
|
199 |
return None, ""
|
200 |
|
|
|
204 |
else:
|
205 |
image = input_image
|
206 |
|
207 |
+
image = image.resize((1280, 720))
|
208 |
+
|
209 |
+
# Detect interactions
|
210 |
+
interactions, pose_results, segmentation_map, depth_map = self.detect_wall_interaction(image)
|
211 |
|
212 |
# Visualize results
|
213 |
+
result_image = self.visualize_results(image, interactions, pose_results)
|
214 |
|
215 |
+
# Create interaction information text
|
216 |
info_text = []
|
217 |
+
for interaction in interactions:
|
218 |
+
info_text.append(f"\nPerson {interaction['person_id'] + 1}:")
|
219 |
+
if interaction["right_hand_touching"]:
|
220 |
+
info_text.append(f"Right hand is touching {interaction['right_hand_touching_object']}")
|
221 |
+
if interaction["left_hand_touching"]:
|
222 |
+
info_text.append(f"Left hand is touching {interaction['left_hand_touching_object']}")
|
223 |
+
info_text.append(f"Right hand distance to wall: {interaction['right_hand_distance']:.2f}")
|
224 |
+
info_text.append(f"Left hand distance to wall: {interaction['left_hand_distance']:.2f}")
|
225 |
+
|
226 |
|
227 |
+
# Add color to segmentation
|
228 |
+
mask = np.zeros((*segmentation_map.shape, 3), dtype=np.uint8)
|
229 |
+
colors = np.random.randint(0, 255, size=(100, 3))
|
|
|
230 |
|
231 |
+
for cl_id in np.unique(segmentation_map):
|
232 |
+
mask_array = np.array(segmentation_map == cl_id)
|
233 |
+
color = colors[cl_id % len(colors)]
|
234 |
+
mask[mask_array] = color
|
|
|
235 |
|
236 |
+
return result_image, mask, depth_map, "\n".join(info_text)
|
237 |
|
238 |
|
239 |
def create_gradio_interface():
|
240 |
"""Create Gradio interface"""
|
241 |
+
detector = InteractionDetector()
|
242 |
|
243 |
with gr.Blocks() as interface:
|
244 |
+
gr.Markdown("# Object Interaction Detection")
|
245 |
+
gr.Markdown("Upload an image to detect when people are touching objects.")
|
|
|
|
|
|
|
246 |
|
247 |
with gr.Row():
|
248 |
with gr.Column():
|
249 |
input_image = gr.Image(label="Input Image")
|
250 |
+
process_button = gr.Button("Detect Interactions")
|
251 |
|
252 |
with gr.Column():
|
253 |
output_image = gr.Image(label="Detection Results")
|
254 |
+
interaction_info = gr.Textbox(
|
255 |
+
label="Interaction Information",
|
256 |
lines=10,
|
257 |
+
placeholder="Interaction details will appear here..."
|
258 |
)
|
259 |
+
segmentation_im = gr.Image(label="Segmentaiton Results")
|
260 |
+
depth_im = gr.Image(label="Depth Results")
|
261 |
|
262 |
process_button.click(
|
263 |
fn=detector.process_image,
|
264 |
inputs=input_image,
|
265 |
+
outputs=[output_image, segmentation_im, depth_im, interaction_info]
|
266 |
)
|
267 |
|
268 |
gr.Examples(
|
269 |
examples=[
|
270 |
+
"https://img.freepik.com/premium-photo/happy-black-man-opening-door-gesturing-okay-approving-new-home_116547-23954.jpg?w=1800",
|
271 |
+
"https://static3.bigstockphoto.com/6/7/2/large1500/276757975.jpg"
|
272 |
],
|
273 |
inputs=input_image
|
274 |
)
|
275 |
|
276 |
return interface
|
277 |
|
278 |
+
interface = create_gradio_interface()
|
279 |
if __name__ == "__main__":
|
280 |
+
interface.launch(debug=True)
|
|