Spaces:
Sleeping
Sleeping
Update main.py
Browse files
main.py
CHANGED
@@ -1,5 +1,5 @@
|
|
1 |
from fastapi import FastAPI, File, UploadFile, HTTPException
|
2 |
-
from pydantic import BaseModel
|
3 |
import base64
|
4 |
import io
|
5 |
import os
|
@@ -24,10 +24,7 @@ yolo_model = get_yolo_model(model_path="weights/best.pt")
|
|
24 |
|
25 |
# Handle device placement
|
26 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
27 |
-
|
28 |
-
yolo_model = yolo_model.cuda()
|
29 |
-
else:
|
30 |
-
yolo_model = yolo_model.cpu()
|
31 |
|
32 |
# Load caption model and processor
|
33 |
try:
|
@@ -38,7 +35,7 @@ try:
|
|
38 |
"weights/icon_caption_florence",
|
39 |
torch_dtype=torch.float16,
|
40 |
trust_remote_code=True,
|
41 |
-
).to("cuda")
|
42 |
except Exception as e:
|
43 |
logger.warning(f"Failed to load caption model on GPU: {e}. Falling back to CPU.")
|
44 |
model = AutoModelForCausalLM.from_pretrained(
|
@@ -48,7 +45,7 @@ except Exception as e:
|
|
48 |
)
|
49 |
|
50 |
caption_model_processor = {"processor": processor, "model": model}
|
51 |
-
logger.info("Finished loading models
|
52 |
|
53 |
# Initialize FastAPI app
|
54 |
app = FastAPI()
|
@@ -56,51 +53,44 @@ app = FastAPI()
|
|
56 |
MAX_QUEUE_SIZE = 10 # Set a reasonable limit based on your system capacity
|
57 |
request_queue = asyncio.Queue(maxsize=MAX_QUEUE_SIZE)
|
58 |
|
59 |
-
# Define
|
60 |
class ProcessResponse(BaseModel):
|
61 |
image: str # Base64 encoded image
|
62 |
parsed_content_list: str
|
63 |
label_coordinates: str
|
64 |
|
65 |
|
66 |
-
#
|
67 |
async def worker():
|
68 |
-
"""
|
69 |
-
Background worker to process tasks from the request queue sequentially.
|
70 |
-
"""
|
71 |
while True:
|
72 |
-
task = await request_queue.get()
|
73 |
try:
|
74 |
-
await task
|
75 |
except Exception as e:
|
76 |
logger.error(f"Error while processing task: {e}")
|
77 |
finally:
|
78 |
-
request_queue.task_done()
|
79 |
|
80 |
|
81 |
-
# Start
|
82 |
@app.on_event("startup")
|
83 |
async def startup_event():
|
84 |
logger.info("Starting background worker...")
|
85 |
-
|
86 |
-
asyncio.create_task(worker()) # Start the worker in the background
|
87 |
|
88 |
|
89 |
-
#
|
90 |
async def process(image_input: Image.Image, box_threshold: float, iou_threshold: float) -> ProcessResponse:
|
91 |
-
"""
|
92 |
-
Asynchronously processes an image using YOLO and caption models.
|
93 |
-
"""
|
94 |
try:
|
95 |
-
# Define
|
96 |
image_save_path = "imgs/saved_image_demo.png"
|
97 |
os.makedirs(os.path.dirname(image_save_path), exist_ok=True)
|
98 |
|
99 |
-
# Save
|
100 |
image_input.save(image_save_path)
|
101 |
logger.debug(f"Image saved to: {image_save_path}")
|
102 |
|
103 |
-
#
|
104 |
box_overlay_ratio = image_input.size[0] / 3200
|
105 |
draw_bbox_config = {
|
106 |
"text_scale": 0.8 * box_overlay_ratio,
|
@@ -152,7 +142,7 @@ async def process(image_input: Image.Image, box_threshold: float, iou_threshold:
|
|
152 |
raise HTTPException(status_code=500, detail=f"Failed to process the image: {e}")
|
153 |
|
154 |
|
155 |
-
#
|
156 |
@app.post("/process_image", response_model=ProcessResponse)
|
157 |
async def process_image(
|
158 |
image_file: UploadFile = File(...),
|
@@ -160,22 +150,22 @@ async def process_image(
|
|
160 |
iou_threshold: float = 0.1,
|
161 |
):
|
162 |
try:
|
163 |
-
# Read
|
164 |
contents = await image_file.read()
|
165 |
try:
|
166 |
image_input = Image.open(io.BytesIO(contents)).convert("RGB")
|
167 |
-
except UnidentifiedImageError
|
168 |
-
logger.error(
|
169 |
raise HTTPException(status_code=400, detail="Unsupported image format.")
|
170 |
|
171 |
-
# Create
|
172 |
task = asyncio.create_task(process(image_input, box_threshold, iou_threshold))
|
173 |
|
174 |
-
# Add
|
175 |
await request_queue.put(task)
|
176 |
logger.info(f"Task added to queue. Current queue size: {request_queue.qsize()}")
|
177 |
|
178 |
-
# Wait for
|
179 |
response = await task
|
180 |
|
181 |
return response
|
@@ -183,4 +173,4 @@ async def process_image(
|
|
183 |
raise he
|
184 |
except Exception as e:
|
185 |
logger.error(f"Error processing image: {e}")
|
186 |
-
raise HTTPException(status_code=500, detail=f"Internal server error: {e}")
|
|
|
1 |
from fastapi import FastAPI, File, UploadFile, HTTPException
|
2 |
+
from pydantic import BaseModel
|
3 |
import base64
|
4 |
import io
|
5 |
import os
|
|
|
24 |
|
25 |
# Handle device placement
|
26 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
27 |
+
yolo_model = yolo_model.to(device)
|
|
|
|
|
|
|
28 |
|
29 |
# Load caption model and processor
|
30 |
try:
|
|
|
35 |
"weights/icon_caption_florence",
|
36 |
torch_dtype=torch.float16,
|
37 |
trust_remote_code=True,
|
38 |
+
).to("cuda" if torch.cuda.is_available() else "cpu")
|
39 |
except Exception as e:
|
40 |
logger.warning(f"Failed to load caption model on GPU: {e}. Falling back to CPU.")
|
41 |
model = AutoModelForCausalLM.from_pretrained(
|
|
|
45 |
)
|
46 |
|
47 |
caption_model_processor = {"processor": processor, "model": model}
|
48 |
+
logger.info("Finished loading models!")
|
49 |
|
50 |
# Initialize FastAPI app
|
51 |
app = FastAPI()
|
|
|
53 |
MAX_QUEUE_SIZE = 10 # Set a reasonable limit based on your system capacity
|
54 |
request_queue = asyncio.Queue(maxsize=MAX_QUEUE_SIZE)
|
55 |
|
56 |
+
# Define response model
|
57 |
class ProcessResponse(BaseModel):
|
58 |
image: str # Base64 encoded image
|
59 |
parsed_content_list: str
|
60 |
label_coordinates: str
|
61 |
|
62 |
|
63 |
+
# Background worker to process queue tasks
|
64 |
async def worker():
|
|
|
|
|
|
|
65 |
while True:
|
66 |
+
task = await request_queue.get()
|
67 |
try:
|
68 |
+
await task
|
69 |
except Exception as e:
|
70 |
logger.error(f"Error while processing task: {e}")
|
71 |
finally:
|
72 |
+
request_queue.task_done()
|
73 |
|
74 |
|
75 |
+
# Start worker on startup
|
76 |
@app.on_event("startup")
|
77 |
async def startup_event():
|
78 |
logger.info("Starting background worker...")
|
79 |
+
asyncio.create_task(worker())
|
|
|
80 |
|
81 |
|
82 |
+
# Image processing function
|
83 |
async def process(image_input: Image.Image, box_threshold: float, iou_threshold: float) -> ProcessResponse:
|
|
|
|
|
|
|
84 |
try:
|
85 |
+
# Define save path
|
86 |
image_save_path = "imgs/saved_image_demo.png"
|
87 |
os.makedirs(os.path.dirname(image_save_path), exist_ok=True)
|
88 |
|
89 |
+
# Save image
|
90 |
image_input.save(image_save_path)
|
91 |
logger.debug(f"Image saved to: {image_save_path}")
|
92 |
|
93 |
+
# YOLO and caption model inference
|
94 |
box_overlay_ratio = image_input.size[0] / 3200
|
95 |
draw_bbox_config = {
|
96 |
"text_scale": 0.8 * box_overlay_ratio,
|
|
|
142 |
raise HTTPException(status_code=500, detail=f"Failed to process the image: {e}")
|
143 |
|
144 |
|
145 |
+
# API endpoint for processing images
|
146 |
@app.post("/process_image", response_model=ProcessResponse)
|
147 |
async def process_image(
|
148 |
image_file: UploadFile = File(...),
|
|
|
150 |
iou_threshold: float = 0.1,
|
151 |
):
|
152 |
try:
|
153 |
+
# Read image file
|
154 |
contents = await image_file.read()
|
155 |
try:
|
156 |
image_input = Image.open(io.BytesIO(contents)).convert("RGB")
|
157 |
+
except UnidentifiedImageError:
|
158 |
+
logger.error("Unsupported image format.")
|
159 |
raise HTTPException(status_code=400, detail="Unsupported image format.")
|
160 |
|
161 |
+
# Create processing task
|
162 |
task = asyncio.create_task(process(image_input, box_threshold, iou_threshold))
|
163 |
|
164 |
+
# Add task to queue
|
165 |
await request_queue.put(task)
|
166 |
logger.info(f"Task added to queue. Current queue size: {request_queue.qsize()}")
|
167 |
|
168 |
+
# Wait for processing to complete
|
169 |
response = await task
|
170 |
|
171 |
return response
|
|
|
173 |
raise he
|
174 |
except Exception as e:
|
175 |
logger.error(f"Error processing image: {e}")
|
176 |
+
raise HTTPException(status_code=500, detail=f"Internal server error: {e}")
|