fix: form with multi-part python, np format for model

app.py

@@ -5,7 +5,7 @@ from pathlib import Path
 from typing import Literal, Optional
 import base64
 
-import cv2
+import numpy as np
 import supervision as sv
 import uvicorn
 import yaml
@@ -24,7 +24,7 @@ PROCESSED_FOLDER = Path(os.environ.get("TEMP_IMG_FOLDER", "temp_dir"))
 PROCESSED_FOLDER.mkdir(parents=True, exist_ok=True)
 BASE_URL = "http://127.0.0.1:8000"
 
-API_PARTNER_KEY = os.environ.get("API_PARTNER_KEY", "")
+API_PARTNER_KEY = os.environ.get("API_PARTNER_KEY", "dev-AfghDgr3fgf74vc")
 API_KEY_NAME = "x-api-key"
 api_key_header = APIKeyHeader(name=API_KEY_NAME, auto_error=False)
 
@@ -38,10 +38,14 @@ app.add_middleware(
     allow_headers=["*"],
 )
 
-pdf_converter: PdfConverter = PdfConverter()
+pdf_converter: PdfConverter = PdfConverter(dpi=150)
 grounding_model = Model(
-    model_config_path=os.environ.get("GROUNDING_DINO_CONFIG"),
-    model_checkpoint_path=os.environ.get("GROUNDING_DINO_CHECKPOINT"),
+    model_config_path=os.environ.get(
+        "GROUNDING_DINO_CONFIG", "groundingdino/config/GroundingDINO_SwinT_OGC.py"
+    ),
+    model_checkpoint_path=os.environ.get(
+        "GROUNDING_DINO_CHECKPOINT", "gdino_checkpoints/groundingdino_swint_ogc.pth"
+    ),
     device="cuda:0",
 )
 BOX_THRESHOLD = 0.4
@@ -112,8 +116,8 @@ async def api_key_middleware(request: Request, call_next):
 
 @app.post("/crop_ooi")
 async def crop_object_of_interest(
-    document_file: Optional[UploadFile] = File(
-        None, description="The document to process."
+    document_file: UploadFile = File(
+        ..., description="The document to process."
     ),
     concept_list: Optional[list[str]] = Form(
         ["ID document"], description="List of concepts to detect e.g. dog, cat, rain"
@@ -144,7 +148,7 @@ async def crop_object_of_interest(
         detection_img_list = []
         for image in images:
             detections, labels = grounding_model.predict_with_caption(
-                image=images,
+                image=np.asarray(image),
                 caption=caption,
                 box_threshold=box_threshold,
                 text_threshold=text_threshold,
@@ -158,7 +162,7 @@ async def crop_object_of_interest(
             ]
             for i, bbox in enumerate(detections.xyxy):
                 x_min, y_min, x_max, y_max = tuple(bbox)
-                patch = image[int(y_min) : int(y_max), int(x_min) : int(x_max)]
+                patch = image.crop((int(y_min), int(y_max), int(x_min), int(x_max)))
                 image_json = output_img(processed_img=patch, output=output)
                 image_list.append(image_json)
 
@@ -179,7 +183,7 @@ async def crop_object_of_interest(
 
     except Exception as e:
         print(f"{e}")
-        return JSONResponse(status_code=500, content=str(e))
+        return JSONResponse(status_code=501, content=str(e))
 
 
 if __name__ == "__main__":