Spaces:
Running
on
Zero
Running
on
Zero
Actually run inference on the image
Browse files
app.py
CHANGED
@@ -12,6 +12,7 @@ import time
|
|
12 |
import uuid
|
13 |
|
14 |
import subprocess
|
|
|
15 |
subprocess.run(
|
16 |
"pip install gradio_rerun-0.23.0a2.tar.gz",
|
17 |
shell=True,
|
@@ -40,11 +41,14 @@ image = Image.open(requests.get(url, stream=True).raw)
|
|
40 |
processor = DetrImageProcessor.from_pretrained("facebook/detr-resnet-50")
|
41 |
model = DetrForObjectDetection.from_pretrained("facebook/detr-resnet-50")
|
42 |
|
|
|
43 |
# Whenever we need a recording, we construct a new recording stream.
|
44 |
# As long as the app and recording IDs remain the same, the data
|
45 |
# will be merged by the Viewer.
|
46 |
def get_recording(recording_id: str) -> rr.RecordingStream:
|
47 |
-
return rr.RecordingStream(
|
|
|
|
|
48 |
|
49 |
|
50 |
# A task can directly log to a binary stream, which is routed to the embedded viewer.
|
@@ -53,7 +57,7 @@ def get_recording(recording_id: str) -> rr.RecordingStream:
|
|
53 |
# This is the preferred way to work with Rerun in Gradio since your data can be immediately and
|
54 |
# incrementally seen by the viewer. Also, there are no ephemeral RRDs to cleanup or manage.
|
55 |
@spaces.GPU
|
56 |
-
def
|
57 |
# Here we get a recording using the provided recording id.
|
58 |
rec = get_recording(recording_id)
|
59 |
stream = rec.binary_stream()
|
@@ -73,108 +77,38 @@ def streaming_repeated_blur(recording_id: str, img):
|
|
73 |
rec.log("image", rr.Image(img))
|
74 |
yield stream.read()
|
75 |
|
76 |
-
|
77 |
-
|
78 |
-
|
79 |
|
80 |
# convert outputs (bounding boxes and class logits) to COCO API
|
81 |
# let's only keep detections with score > 0.9
|
82 |
-
target_sizes = torch.tensor([image.size[::-1]])
|
83 |
-
results = processor.post_process_object_detection(outputs, target_sizes=target_sizes, threshold=0.9)[0]
|
84 |
-
|
85 |
-
print(results)
|
86 |
-
rec.log("image/objects", rr.Boxes2D(sizes=results["boxes"], labels=[model.config.id2label[label.item()] for label in results["labels"]]))
|
87 |
-
|
88 |
-
# Ensure we consume everything from the recording.
|
89 |
-
stream.flush()
|
90 |
-
yield stream.read()
|
91 |
-
|
92 |
-
|
93 |
-
# In this example the user is able to add keypoints to an image visualized in Rerun.
|
94 |
-
# These keypoints are stored in the global state, we use the session id to keep track of which keypoints belong
|
95 |
-
# to a specific session (https://www.gradio.app/guides/state-in-blocks).
|
96 |
-
#
|
97 |
-
# The current session can be obtained by adding a parameter of type `gradio.Request` to your event listener functions.
|
98 |
-
Keypoint = tuple[float, float]
|
99 |
-
keypoints_per_session_per_sequence_index: dict[str, dict[int, list[Keypoint]]] = {}
|
100 |
-
|
101 |
-
|
102 |
-
def get_keypoints_for_user_at_sequence_index(request: gr.Request, sequence: int) -> list[Keypoint]:
|
103 |
-
per_sequence = keypoints_per_session_per_sequence_index[request.session_hash]
|
104 |
-
if sequence not in per_sequence:
|
105 |
-
per_sequence[sequence] = []
|
106 |
-
|
107 |
-
return per_sequence[sequence]
|
108 |
-
|
109 |
-
|
110 |
-
def initialize_instance(request: gr.Request) -> None:
|
111 |
-
keypoints_per_session_per_sequence_index[request.session_hash] = {}
|
112 |
|
|
|
|
|
|
|
|
|
113 |
|
114 |
-
|
115 |
-
|
116 |
-
|
117 |
-
|
118 |
-
|
119 |
-
|
120 |
-
|
121 |
-
|
122 |
-
|
123 |
-
# `gr.Request`: https://www.gradio.app/main/docs/gradio/request
|
124 |
-
def register_keypoint(
|
125 |
-
active_recording_id: str,
|
126 |
-
current_timeline: str,
|
127 |
-
current_time: float,
|
128 |
-
request: gr.Request,
|
129 |
-
evt: SelectionChange,
|
130 |
-
):
|
131 |
-
if active_recording_id == "":
|
132 |
-
return
|
133 |
-
|
134 |
-
if current_timeline != "iteration":
|
135 |
-
return
|
136 |
-
|
137 |
-
# We can only log a keypoint if the user selected only a single item.
|
138 |
-
if len(evt.items) != 1:
|
139 |
-
return
|
140 |
-
item = evt.items[0]
|
141 |
-
|
142 |
-
# If the selected item isn't an entity, or we don't have its position, then bail out.
|
143 |
-
if item.kind != "entity" or item.position is None:
|
144 |
-
return
|
145 |
-
|
146 |
-
# Now we can produce a valid keypoint.
|
147 |
-
rec = get_recording(active_recording_id)
|
148 |
-
stream = rec.binary_stream()
|
149 |
-
|
150 |
-
# We round `current_time` toward 0, because that gives us the sequence index
|
151 |
-
# that the user is currently looking at, due to the Viewer's latest-at semantics.
|
152 |
-
index = math.floor(current_time)
|
153 |
-
|
154 |
-
# We keep track of the keypoints per sequence index for each user manually.
|
155 |
-
keypoints = get_keypoints_for_user_at_sequence_index(request, index)
|
156 |
-
keypoints.append(item.position[0:2])
|
157 |
-
|
158 |
-
rec.set_time("iteration", sequence=index)
|
159 |
-
rec.log(f"{item.entity_path}/keypoint", rr.Points2D(keypoints, radii=2))
|
160 |
|
161 |
# Ensure we consume everything from the recording.
|
162 |
stream.flush()
|
163 |
yield stream.read()
|
164 |
|
165 |
|
166 |
-
def track_current_time(evt: TimeUpdate):
|
167 |
-
return evt.time
|
168 |
-
|
169 |
-
|
170 |
-
def track_current_timeline_and_time(evt: TimelineChange):
|
171 |
-
return evt.timeline, evt.time
|
172 |
-
|
173 |
with gr.Blocks() as demo:
|
174 |
with gr.Row():
|
175 |
img = gr.Image(interactive=True, label="Image")
|
176 |
with gr.Column():
|
177 |
-
|
178 |
|
179 |
with gr.Row():
|
180 |
viewer = Rerun(
|
@@ -195,20 +129,11 @@ with gr.Blocks() as demo:
|
|
195 |
|
196 |
# When registering the event listeners, we pass the `recording_id` in as input in order to create
|
197 |
# a recording stream using that id.
|
198 |
-
|
199 |
# Using the `viewer` as an output allows us to stream data to it by yielding bytes from the callback.
|
200 |
-
|
201 |
inputs=[recording_id, img],
|
202 |
outputs=[viewer],
|
203 |
)
|
204 |
-
viewer.selection_change(
|
205 |
-
register_keypoint,
|
206 |
-
inputs=[recording_id, current_timeline, current_time],
|
207 |
-
outputs=[viewer],
|
208 |
-
)
|
209 |
-
viewer.time_update(track_current_time, outputs=[current_time])
|
210 |
-
viewer.timeline_change(track_current_timeline_and_time, outputs=[current_timeline, current_time])
|
211 |
-
|
212 |
-
|
213 |
if __name__ == "__main__":
|
214 |
-
demo.launch()
|
|
|
12 |
import uuid
|
13 |
|
14 |
import subprocess
|
15 |
+
|
16 |
subprocess.run(
|
17 |
"pip install gradio_rerun-0.23.0a2.tar.gz",
|
18 |
shell=True,
|
|
|
41 |
processor = DetrImageProcessor.from_pretrained("facebook/detr-resnet-50")
|
42 |
model = DetrForObjectDetection.from_pretrained("facebook/detr-resnet-50")
|
43 |
|
44 |
+
|
45 |
# Whenever we need a recording, we construct a new recording stream.
|
46 |
# As long as the app and recording IDs remain the same, the data
|
47 |
# will be merged by the Viewer.
|
48 |
def get_recording(recording_id: str) -> rr.RecordingStream:
|
49 |
+
return rr.RecordingStream(
|
50 |
+
application_id="rerun_example_gradio", recording_id=recording_id
|
51 |
+
)
|
52 |
|
53 |
|
54 |
# A task can directly log to a binary stream, which is routed to the embedded viewer.
|
|
|
57 |
# This is the preferred way to work with Rerun in Gradio since your data can be immediately and
|
58 |
# incrementally seen by the viewer. Also, there are no ephemeral RRDs to cleanup or manage.
|
59 |
@spaces.GPU
|
60 |
+
def streaming_object_detection(recording_id: str, img):
|
61 |
# Here we get a recording using the provided recording id.
|
62 |
rec = get_recording(recording_id)
|
63 |
stream = rec.binary_stream()
|
|
|
77 |
rec.log("image", rr.Image(img))
|
78 |
yield stream.read()
|
79 |
|
80 |
+
with torch.inference_mode():
|
81 |
+
inputs = processor(images=img, return_tensors="pt")
|
82 |
+
outputs = model(**inputs)
|
83 |
|
84 |
# convert outputs (bounding boxes and class logits) to COCO API
|
85 |
# let's only keep detections with score > 0.9
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
86 |
|
87 |
+
target_sizes = torch.tensor([img.height, img.width])
|
88 |
+
results = processor.post_process_object_detection(
|
89 |
+
outputs, target_sizes=target_sizes, threshold=0.9
|
90 |
+
)[0]
|
91 |
|
92 |
+
print(results)
|
93 |
+
rec.log(
|
94 |
+
"image/objects",
|
95 |
+
rr.Boxes2D(
|
96 |
+
array=results["boxes"],
|
97 |
+
array_format=rr.Box2DFormat.XYXY,
|
98 |
+
labels=[model.config.id2label[label.item()] for label in results["labels"]],
|
99 |
+
),
|
100 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
101 |
|
102 |
# Ensure we consume everything from the recording.
|
103 |
stream.flush()
|
104 |
yield stream.read()
|
105 |
|
106 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
107 |
with gr.Blocks() as demo:
|
108 |
with gr.Row():
|
109 |
img = gr.Image(interactive=True, label="Image")
|
110 |
with gr.Column():
|
111 |
+
detect_objects = gr.Button("Detect objects")
|
112 |
|
113 |
with gr.Row():
|
114 |
viewer = Rerun(
|
|
|
129 |
|
130 |
# When registering the event listeners, we pass the `recording_id` in as input in order to create
|
131 |
# a recording stream using that id.
|
132 |
+
detect_objects.click(
|
133 |
# Using the `viewer` as an output allows us to stream data to it by yielding bytes from the callback.
|
134 |
+
streaming_object_detection,
|
135 |
inputs=[recording_id, img],
|
136 |
outputs=[viewer],
|
137 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
138 |
if __name__ == "__main__":
|
139 |
+
demo.launch(ssr_mode=False)
|