Use docker

.dockerignore

@@ -0,0 +1 @@
+CutLER/

Dockerfile

@@ -0,0 +1,19 @@
+FROM python:3.9
+
+WORKDIR /code
+
+# Install git
+RUN apt-get update && apt-get install -y git
+
+# Install CutLER
+RUN git clone --recursive https://github.com/Ad-Visual/CutLER
+
+RUN pip install --no-cache-dir --upgrade -r /code/CutLER/requirements.txt
+
+COPY ./requirements.txt /code/requirements.txt
+
+RUN pip install --no-cache-dir --upgrade -r /code/requirements.txt
+
+COPY . /code
+
+CMD ["uvicorn", "app.main:app", "--host", "0.0.0.0", "--port", "7860"]

README.md

@@ -3,7 +3,7 @@ title: MaskCut
 emoji: 🌍
 colorFrom: pink
 colorTo: pink
-sdk: gradio
+sdk: docker
 sdk_version: 3.16.2
 app_file: app.py
 pinned: false

predict.py

@@ -17,7 +17,7 @@ def predict(package, image_base64: str, threshold: float, num_objects: int):
 
     # We copy the image that and fill it with black, to get the dimensions
     rgb = np.array(canvas)
-    model = package.get('model')
+    model : Model = package.get('model')
     masks = model(image, threshold, num_objects)
 
     for mask in masks:  

requirements.txt

@@ -1,12 +1,12 @@
 git+https://github.com/lucasb-eyer/pydensecrf
-gradio==3.16.2
-numpy==1.23.5
+torch==1.8.1
+torchvision==0.9.1
+numpy==1.24.2
 opencv-python==4.6.0.66
-pycocotools==2.0.6
 scikit-image==0.19.2
-torch==1.13.1
-torchvision==0.14.1
 tqdm==4.64.1
+
+pycocotools==2.0.6
 fastapi==0.94.0
 pydantic==1.8.2
 uvicorn==0.21.0

server.py

@@ -17,9 +17,6 @@ import torch
 import os
 import sys
 
-# Server Framework
-import uvicorn
-
 app = FastAPI(
     title="AdVisual MaskCut Model",
     description="Description of the ML Model",
@@ -136,8 +133,4 @@ async def websocket_endpoint(websocket: WebSocket):
 
           await websocket.close()
         except WebSocketDisconnect:
-          break 
-
-if __name__ == '__main__':
-    # server api
-    uvicorn.run("main:app", host="0.0.0.0", port=8000, reload=True)
+          break