Update pipeline/logic.py
Browse files- pipeline/logic.py +10 -1
pipeline/logic.py
CHANGED
|
@@ -24,6 +24,8 @@ def download_image_from_url(image_url: str) -> Image.Image:
|
|
| 24 |
print(" [Download] ✅ Image downloaded and standardized to RGB.")
|
| 25 |
return image_rgb
|
| 26 |
|
|
|
|
|
|
|
| 27 |
def detect_and_crop(image: Image.Image, object_name: str, models: dict) -> Image.Image:
|
| 28 |
print(f"\n [Detect & Crop] Starting detection for object: '{object_name}'")
|
| 29 |
image_np = np.array(image.convert("RGB"))
|
|
@@ -32,9 +34,16 @@ def detect_and_crop(image: Image.Image, object_name: str, models: dict) -> Image
|
|
| 32 |
inputs = models['processor_gnd'](images=image, text=prompt, return_tensors="pt").to(models['device'])
|
| 33 |
with torch.no_grad():
|
| 34 |
outputs = models['model_gnd'](**inputs)
|
|
|
|
|
|
|
| 35 |
results = models['processor_gnd'].post_process_grounded_object_detection(
|
| 36 |
-
outputs,
|
|
|
|
|
|
|
|
|
|
|
|
|
| 37 |
)
|
|
|
|
| 38 |
if not results or len(results[0]['boxes']) == 0:
|
| 39 |
print(" [Detect & Crop] ⚠ Warning: Grounding DINO did not detect the object. Using full image.")
|
| 40 |
return image
|
|
|
|
| 24 |
print(" [Download] ✅ Image downloaded and standardized to RGB.")
|
| 25 |
return image_rgb
|
| 26 |
|
| 27 |
+
# In pipeline/logic.py
|
| 28 |
+
|
| 29 |
def detect_and_crop(image: Image.Image, object_name: str, models: dict) -> Image.Image:
|
| 30 |
print(f"\n [Detect & Crop] Starting detection for object: '{object_name}'")
|
| 31 |
image_np = np.array(image.convert("RGB"))
|
|
|
|
| 34 |
inputs = models['processor_gnd'](images=image, text=prompt, return_tensors="pt").to(models['device'])
|
| 35 |
with torch.no_grad():
|
| 36 |
outputs = models['model_gnd'](**inputs)
|
| 37 |
+
|
| 38 |
+
# THIS IS THE CORRECTED LINE:
|
| 39 |
results = models['processor_gnd'].post_process_grounded_object_detection(
|
| 40 |
+
outputs,
|
| 41 |
+
inputs.input_ids,
|
| 42 |
+
object_detection_threshold=0.4, # Renamed from box_threshold
|
| 43 |
+
text_detection_threshold=0.3, # Renamed from text_threshold
|
| 44 |
+
target_sizes=[(height, width)]
|
| 45 |
)
|
| 46 |
+
|
| 47 |
if not results or len(results[0]['boxes']) == 0:
|
| 48 |
print(" [Detect & Crop] ⚠ Warning: Grounding DINO did not detect the object. Using full image.")
|
| 49 |
return image
|