Imageai: Image Detection File stream and numpy

[acleitao76]

Can you provide some example using those input_types to the detectObjectsFromImage method? I have a PIL image and I'm having a hard time converting it to File Stream or numpy array. Look all over the internet and still not working. Amazing library by the way. I got really impressed how good it is after training my own yolo model

[acleitao76]

this is how I implemented it's a very simple code but I'm still getting this error:
raise ValueError("Tensor %s is not an element of this graph." % obj)
ValueError: Tensor Tensor("last1/BiasAdd:0", shape=(?, ?, ?, 18), dtype=float32) is not an element of this graph.
I0815 07:51:06.359514 123145486721024 _internal.py:122] 127.0.0.1 - - [15/Aug/2019 07:51:06] "POST /detectbib HTTP/1.1" 500 -
Is it me or there is something wrong?

@app.route('/detectbib', methods=['POST'])
def detectorBib():
    req = request.get_json()
    imgdata = base64.b64decode(req["image"])
    imgTemp =Image.open(io.BytesIO(imgdata))
    img=cv2.cvtColor(np.array(imgTemp), cv2.COLOR_BGR2RGB)
    bibs = bibDetector(img)

def bibDetector(image):
    returned_image, detections = detector.detectObjectsFromImage(input_type="array", 
                                                                 input_image=image, output_type="array",
                                                                 minimum_percentage_probability=30)
    return returned_image

I'm just trying to get to work without having to save every image in the hard disk. I want process 5000 of my own images and saving to the hard disk will generate a lot o IO. If you have any examples on how this should be done(I get the image in a base64 format as a parameter). The Idea is to call this service from my image processing software and identify bibs and badges... as the head of the person

[acleitao76]

Found my answer. Keras has some trouble working multithreaded flask is multithread as I'm working command line I just added to flask run --without-threads. There is a way to fix it inside imageai. I will do it as soon as I have the time and submit. For now ... just disable threading will solve it