Tinyyolov3's test data is wrong

[snnn]

2019-11-08 12:48:51.317577463 [E:onnxruntime:Default, runner.cc:479 RunTaskImpl] tinyyolov3:output=yolonms_layer_1:2:expect tensor(float) got tensor(int32)
2019-11-08 12:48:51.317648314 [E:onnxruntime:Default, runner.h:76 finish] tinyyolov3: type mismatch. Dataset:/data/testdata/m/opset11/tinyyolov3/test_data_set_0

[snnn]

microsoft/onnxruntime#2356

[SiR0N]

Hi, I also face this problem with the model at running:
sess = rt.InferenceSession("yolov3tiny.onnx")

error:

 self._sess.load_model(path_or_bytes)
RuntimeError: [ONNXRuntimeError] : 10 : INVALID_GRAPH : Load model from yolov3tiny.onnx failed:Node:Resize Node (Resize) has input size 3 not in range [min=2, max=2].

[EmmaNingMS]

@jiafatom

[jiafatom]

Updated the test data and checked in here

[jiafatom]

@SiR0N you are using the old onnxruntime, the error msg shows the number of input for Resize is 2, however, opset 11 has changed this. Please use onnxruntime 1.0+ to test.

[snnn]

Hi @SiR0N
Which onnxruntime version were you using?

[SiR0N]

hi, @jiafatom and @snnn I upgrade it to 1.0.0 and now it "works" (I used 0.4.0 before)

Now I face another problem, it is related to the inputs.

image_data = preprocess(image)
image_size = np.array([image.size[1], image.size[0]], dtype=np.float).reshape(1, 2)

def inference(sess, preprocessed_image, image_size):
 
    input_name = sess.get_inputs()[0].name
    input_name2 = sess.get_inputs()[1].name
    output_name0 = sess.get_outputs()[0].name
    output_name1 = sess.get_outputs()[1].name
    output_name2 = sess.get_outputs()[2].name


    boxes, scores, indices = sess.run([output_name0, output_name1, output_name2],
                                      {input_name: preprocessed_image, input_name2: image_size}) #error here

    return boxes, scores, indices

print(rt.__version__)
sess = rt.InferenceSession("yolov3tiny.onnx")


boxes, scores, indices = inference(sess, image_data, image_size)

Error:

return self._sess.run(output_names, input_feed, run_options)
onnxruntime.capi.onnxruntime_pybind11_state.InvalidArgument: [ONNXRuntimeError] : 2 : INVALID_ARGUMENT : Unexpected input data type. Actual: (class onnxruntime::NonOnnxType<double>) , expected: (class onnxruntime::NonOnnxType<float>)

[jiafatom]

This error means that it needs a input of type float, but the data you provided is of type double, can you convert your data?

[SiR0N]

I tried that and it did not work

boxes, scores, indices = sess.run([output_name0, output_name1, output_name2],
                                      {input_name: preprocessed_image.astype(float), input_name2: image_size.astype(float)})

and they are already float:

image = Image.open('dog.jpg')

image_data = preprocess(image)
image_size = np.array([image.size[1], image.size[0]], dtype=np.float).reshape(1, 2)

def preprocess(img):
    model_image_size = (416, 416)
    boxed_image = letterbox_image(img, tuple(reversed(model_image_size)))
    image_data = np.array(boxed_image, dtype='float32')
    image_data /= 255.
    image_data = np.transpose(image_data, [2, 0, 1])
    image_data = np.expand_dims(image_data, 0)
    
    return image_data

[jiafatom]

Here is an example that we run every day and it works fine. Choose is_tiny_yolo=True