Spaces:
Paused
Paused
Update app.py
Browse files
app.py
CHANGED
@@ -20,6 +20,8 @@ from PIL import Image
|
|
20 |
from functools import partial
|
21 |
import io
|
22 |
|
|
|
|
|
23 |
|
24 |
subprocess.run(shlex.split('pip install wheel/torchmcubes-0.1.0-cp310-cp310-linux_x86_64.whl'))
|
25 |
|
@@ -34,6 +36,9 @@ if torch.cuda.is_available():
|
|
34 |
else:
|
35 |
device = "cpu"
|
36 |
|
|
|
|
|
|
|
37 |
model = TSR.from_pretrained(
|
38 |
"stabilityai/TripoSR",
|
39 |
config_name="config.yaml",
|
@@ -47,17 +52,30 @@ ACCESS = os.getenv("ACCESS")
|
|
47 |
SECRET = os.getenv("SECRET")
|
48 |
bedrock = boto3.client(service_name='bedrock', aws_access_key_id = ACCESS, aws_secret_access_key = SECRET, region_name='us-east-1')
|
49 |
bedrock_runtime = boto3.client(service_name='bedrock-runtime', aws_access_key_id = ACCESS, aws_secret_access_key = SECRET, region_name='us-east-1')
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
# {'text': """Blurry, out of frame, out of focus, Detailed, dull, duplicate, bad quality, low resolution, cropped""", 'weight': -1}],
|
54 |
-
# 'cfg_scale': 7, 'seed': 0, 'samples': 1}
|
55 |
-
# request_body = json.dumps(parameters)
|
56 |
-
# response = bedrock_runtime.invoke_model(body=request_body,modelId = 'stability.stable-diffusion-xl-v1')
|
57 |
-
# response_body = json.loads(response.get('body').read())
|
58 |
-
# base64_image_data = base64.b64decode(response_body['artifacts'][0]['base64'])
|
59 |
|
60 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
61 |
|
62 |
|
63 |
def gen_pos_prompt(text):
|
@@ -164,88 +182,47 @@ def generate(image, mc_resolution, formats=["obj", "glb"]):
|
|
164 |
return mesh_path_obj.name, mesh_path_glb.name
|
165 |
|
166 |
def run_example(text_prompt,seed ,do_remove_background, foreground_ratio, mc_resolution):
|
167 |
-
# Step 1: Generate the image from text prompt
|
168 |
image_pil = generate_image_from_text(text_prompt, seed)
|
169 |
-
|
170 |
-
# Step 2: Preprocess the image
|
171 |
preprocessed = preprocess(image_pil, do_remove_background, foreground_ratio)
|
172 |
|
173 |
-
# Step 3: Generate the 3D model
|
174 |
mesh_name_obj, mesh_name_glb = generate(preprocessed, mc_resolution, ["obj", "glb"])
|
175 |
|
176 |
return preprocessed, mesh_name_obj, mesh_name_glb
|
177 |
|
178 |
-
|
179 |
-
|
180 |
-
|
181 |
-
|
182 |
-
|
183 |
-
|
184 |
-
|
185 |
-
|
186 |
-
|
187 |
-
|
188 |
-
|
189 |
-
|
190 |
-
|
191 |
-
|
192 |
-
|
193 |
-
|
194 |
-
|
195 |
-
|
196 |
-
|
197 |
-
|
198 |
-
|
199 |
-
|
200 |
-
|
201 |
-
|
202 |
-
|
203 |
-
|
204 |
-
|
205 |
-
|
206 |
-
|
207 |
-
|
208 |
-
|
209 |
-
|
210 |
-
|
211 |
-
|
212 |
-
|
213 |
-
|
214 |
-
step=32
|
215 |
-
)
|
216 |
-
with gr.Row():
|
217 |
-
submit = gr.Button("Generate", elem_id="generate", variant="primary")
|
218 |
-
with gr.Column():
|
219 |
-
with gr.Tab("OBJ"):
|
220 |
-
output_model_obj = gr.Model3D(
|
221 |
-
label="Output Model (OBJ Format)",
|
222 |
-
interactive=False,
|
223 |
-
)
|
224 |
-
gr.Markdown("Note: Downloaded object will be flipped in case of .obj export. Export .glb instead or manually flip it before usage.")
|
225 |
-
with gr.Tab("GLB"):
|
226 |
-
output_model_glb = gr.Model3D(
|
227 |
-
label="Output Model (GLB Format)",
|
228 |
-
interactive=False,
|
229 |
-
)
|
230 |
-
gr.Markdown("Note: The model shown here has a darker appearance. Download to get correct results.")
|
231 |
-
# with gr.Row(variant="panel"):
|
232 |
-
# gr.Examples(
|
233 |
-
# examples=[
|
234 |
-
# os.path.join("examples", img_name) for img_name in sorted(os.listdir("examples"))
|
235 |
-
# ],
|
236 |
-
# inputs=[text_prompt],
|
237 |
-
# outputs=[processed_image, output_model_obj, output_model_glb],
|
238 |
-
# cache_examples=True,
|
239 |
-
# fn=partial(run_example, do_remove_background=True, foreground_ratio=0.85, mc_resolution=256),
|
240 |
-
# label="Examples",
|
241 |
-
# examples_per_page=20
|
242 |
-
# )
|
243 |
-
submit.click(fn=check_input_image, inputs=[text_prompt]).success(
|
244 |
-
fn=run_example,
|
245 |
-
inputs=[text_prompt, seed, do_remove_background, foreground_ratio, mc_resolution],
|
246 |
-
outputs=[processed_image, output_model_obj, output_model_glb],
|
247 |
-
# outputs=[output_model_obj, output_model_glb],
|
248 |
-
)
|
249 |
-
|
250 |
-
demo.queue(max_size=10)
|
251 |
-
demo.launch(auth=(os.getenv('USERNAME'), os.getenv('PASSWORD')))
|
|
|
20 |
from functools import partial
|
21 |
import io
|
22 |
|
23 |
+
app = FastAPI()
|
24 |
+
|
25 |
|
26 |
subprocess.run(shlex.split('pip install wheel/torchmcubes-0.1.0-cp310-cp310-linux_x86_64.whl'))
|
27 |
|
|
|
36 |
else:
|
37 |
device = "cpu"
|
38 |
|
39 |
+
torch.cuda.synchronize()
|
40 |
+
|
41 |
+
|
42 |
model = TSR.from_pretrained(
|
43 |
"stabilityai/TripoSR",
|
44 |
config_name="config.yaml",
|
|
|
52 |
SECRET = os.getenv("SECRET")
|
53 |
bedrock = boto3.client(service_name='bedrock', aws_access_key_id = ACCESS, aws_secret_access_key = SECRET, region_name='us-east-1')
|
54 |
bedrock_runtime = boto3.client(service_name='bedrock-runtime', aws_access_key_id = ACCESS, aws_secret_access_key = SECRET, region_name='us-east-1')
|
55 |
+
|
56 |
+
def upload_file_to_s3(file_path, bucket_name, object_name=None):
|
57 |
+
s3_client = boto3.client('s3',aws_access_key_id = ACCESS, aws_secret_access_key = SECRET, region_name='us-east-1')
|
|
|
|
|
|
|
|
|
|
|
|
|
58 |
|
59 |
+
if object_name is None:
|
60 |
+
object_name = file_path
|
61 |
+
|
62 |
+
try:
|
63 |
+
s3_client.upload_file(file_path, bucket_name, object_name)
|
64 |
+
except FileNotFoundError:
|
65 |
+
print(f"The file {file_path} was not found.")
|
66 |
+
return False
|
67 |
+
except NoCredentialsError:
|
68 |
+
print("Credentials not available.")
|
69 |
+
return False
|
70 |
+
except PartialCredentialsError:
|
71 |
+
print("Incomplete credentials provided.")
|
72 |
+
return False
|
73 |
+
except Exception as e:
|
74 |
+
print(f"An error occurred: {e}")
|
75 |
+
return False
|
76 |
+
|
77 |
+
print(f"File {file_path} uploaded successfully to {bucket_name}/{object_name}.")
|
78 |
+
return True
|
79 |
|
80 |
|
81 |
def gen_pos_prompt(text):
|
|
|
182 |
return mesh_path_obj.name, mesh_path_glb.name
|
183 |
|
184 |
def run_example(text_prompt,seed ,do_remove_background, foreground_ratio, mc_resolution):
|
|
|
185 |
image_pil = generate_image_from_text(text_prompt, seed)
|
186 |
+
|
|
|
187 |
preprocessed = preprocess(image_pil, do_remove_background, foreground_ratio)
|
188 |
|
|
|
189 |
mesh_name_obj, mesh_name_glb = generate(preprocessed, mc_resolution, ["obj", "glb"])
|
190 |
|
191 |
return preprocessed, mesh_name_obj, mesh_name_glb
|
192 |
|
193 |
+
|
194 |
+
@app.post("/process_text/")
|
195 |
+
async def process_image(
|
196 |
+
text_prompt: str = Form(...),
|
197 |
+
seed: int = Form(...),
|
198 |
+
do_remove_background: bool = Form(...),
|
199 |
+
foreground_ratio: float = Form(...),
|
200 |
+
mc_resolution: int = Form(...),
|
201 |
+
auth: str = Form(...)
|
202 |
+
):
|
203 |
+
|
204 |
+
if auth == os.getenv("AUTHORIZE"):
|
205 |
+
|
206 |
+
preprocessed, mesh_name_obj, mesh_name_glb = run_example(text_prompt,seed ,do_remove_background, foreground_ratio, mc_resolution)
|
207 |
+
# preprocessed = preprocess(image_pil, do_remove_background, foreground_ratio)
|
208 |
+
# mesh_name_obj, mesh_name_glb = generate(preprocessed, mc_resolution)
|
209 |
+
timestamp = datetime.datetime.now().strftime('%Y%m%d%H%M%S%f')
|
210 |
+
object_name = f'object_{timestamp}_1.obj'
|
211 |
+
object_name_2 = f'object_{timestamp}_2.glb'
|
212 |
+
|
213 |
+
if upload_file_to_s3(mesh_name_obj, 'framebucket3d',object_name) and upload_file_to_s3(mesh_name_glb, 'framebucket3d',object_name_2):
|
214 |
+
|
215 |
+
return {
|
216 |
+
"obj_path": f"https://framebucket3d.s3.amazonaws.com/{object_name}",
|
217 |
+
"glb_path": f"https://framebucket3d.s3.amazonaws.com/{object_name_2}"
|
218 |
+
|
219 |
+
}
|
220 |
+
|
221 |
+
else:
|
222 |
+
return {"Internal Server Error": False}
|
223 |
+
else:
|
224 |
+
return {"Authentication":"Failed"}
|
225 |
+
|
226 |
+
if __name__ == "__main__":
|
227 |
+
import uvicorn
|
228 |
+
uvicorn.run(app, host="0.0.0.0", port=7860)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|