def extract_bounding_boxes()

in detection-function/src/run.py [0:0]

• 57 lines of code
• 13 McCabe index (conditional complexity)

def extract_bounding_boxes(image):
    response = sm.invoke_endpoint(
        EndpointName=SAGEMAKER_SSD_ENDPOINT,
        Body=image,
        ContentType='image/jpeg'
    )

    result = json.loads(response['Body'].read())

    # list for bounding boxes
    bounding_boxes = []

    # Remove bounding boxes above the confidence ssd treshold 
    for prediction in result["prediction"]:
        if prediction[1] >= CONFIDENCE_THRESHOLD_SSD:
            bounding_boxes.append(prediction)

    print("Bounding box quantity before NMS: {}".format(len(bounding_boxes)))
    
    i = 0
    while i < len(bounding_boxes):
        boundbox = bounding_boxes[i]
        len_bounding_boxes = len(bounding_boxes)
        i+=1
            
        if boundbox[5] - boundbox[3] < SMALL_BOX_FILTER_SCORE:
            print("Size Removing, Y small: " + str(boundbox[5] - boundbox[3]))
            bounding_boxes.remove(boundbox)
            i=0
            continue

        if boundbox[4] - boundbox[2] < SMALL_BOX_FILTER_SCORE:
            print("Size Removing, x small: " + str(boundbox[4] - boundbox[2]))
            bounding_boxes.remove(boundbox)
            i=0
            continue   
        
        count = 0
        len_bounding_boxes = len(bounding_boxes)
        while count < len_bounding_boxes:
            
            if boundbox != bounding_boxes[count]:
                check = get_iou(boundbox[2:],bounding_boxes[count][2:])
                if check >= CONFIDENCE_THRESHOLD_IOU:        
                    # Best Score
                    if boundbox[1] < bounding_boxes[count][1]:
                        print("Removing: " + str(boundbox[1]))
                        bounding_boxes.remove(boundbox)

                    else:
                        print("Removing: " + str(bounding_boxes[count][1]))
                        bounding_boxes.remove(bounding_boxes[count])
                
                    print("Item Deleted: {}".format(count))
                    len_bounding_boxes -= 1
            
            # Raise Counter
            count += 1
    
    print("Bounding box quantity after NMS: {}".format(len(bounding_boxes)))

    items = []

    for p in bounding_boxes:
         (klass, score, x0, y0, x1, y1) = p

         if score < CONFIDENCE_THRESHOLD_SSD:
            continue
            
         photo = Image.open(io.BytesIO(image))
         shape = OBJECT_CATEGORIES[int(klass)]
         items.append({
            'product_id': shape,
            'object_detection_score': score,
            'position': get_object_boundary_box(photo, (x0, y0, x1, y1)),
        })
        
    print('extract_bounding_boxes: {}'.format(items))
    
    return items